University of Southern California Dissertations and Theses

Alcohol treatment entry and refusal in a sample of older veterans

(USC Thesis Other)

Alcohol treatment entry and refusal in a sample of older veterans

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ALCOHOL TREATMENT ENTRY AND REFUSAL IN A SAMPLE OF
OLDER VETERANS
Copyright 2001
by
Derek Satre
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
in Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)
May 2001
Derek Satre
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
U M I Number: 3027775
___ ®
UMI
UMI Microform 3027775
Copyright 2002 by Bell & Howell Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
Bell & Howell Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY OF SOUTHERN CALIFORNIA
The Graduate School
University Park
LOS ANGELES, CALIFORNIA 900894695
This dissertation, w ritten b y
tkrtK S _________
Under th e direction o f h. D issertation
Com m ittee, and approved b y a ll its mem bers,
has been p resen ted to an d accepted b y The
Graduate School, in p a rtia l fulfillm ent o f
requirem ents fo r th e degree o f
DOCTOR OF PHILOSOPHY
Dean o f Graduate Studies
D a te M ay 11. 2001 ........
DISSERTA TION COMMITTEE
Chairperson
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ii
Table of Contents
Section Page
Table of Contents ii
List of Tables iii
List of Figures iv
Abstract v
Background 1
Hypotheses 24
Method 25
Results 32
Discussion 78
References 99
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
iii
List of Tables
Table Page
Table 1. Demographic Composition of the Stage 1 Sample 35
Table 2. Differences between African Americans and Whites at Stage 1 42
Table 3. Demographics and Severity in Relation to Treatment Interest at 44
Stage 1
Table 4. Summary of Hierarchical Regression Analysis for Variables 46
Predicting Expressed Interest in Substance Abuse Treatment at Stage 1
Table 5. Demographic Composition of Individuals who Participated 51
in Evaluation
Table 6. Age and CAGE scores of those who Participated in Evaluation 53
in Relation to Expression of Interest in Treatment
Table 7. Demographics and Severity in Relation to Evaluation at Stage 2 56
Table 8. Summary of Hierarchical Regression Analysis for Variables 58
Predicting Participation in Evaluation
Table 9. Correlation of Scale Scores 62
Table 10. Scores on the Folstein Mini-Mental Status Exam, in Relation 68
to Published Norms
Table 11. Demographics in Relation to Treatment Entry 69
Table 12. Substance Abuse Indicators in Relation to Treatment Enrollment 70
at Stage 3
Table 13. Anxiety, Depression, and Cognitive Function in Relation to 71
Treatment Enrollment at Stage 3
Table 14. Summary of Hierarchical Regression Analysis for Variables 73
Predicting Treatment Entry
Table 15. Summary of Hierarchical Regression Analysis for CAGE and 75
HRS-D in Predicting Enrollment in Treatment.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Figures
Figure Page
Figure 1. The Behavioral Model 3
Figure 2. The Proposed Developmental Distress Model 4
Figure 3. Destination of Individuals Screened for Treatment 27
Figure 4. Age Distribution of the Sample at Stage 1 35
Figure 5. The Relationship of mean CAGE score to Drinking Frequency 37
Figure 6. Mean Age of Participants by CAGE score 39
Figure 7. Age ofParticipants by Drinking Frequency 40
Figure 8. Path Analysis for Stage 1 Results, Showing Standardized
Regression Coefficients
47
Figure 9. Path Analysis for Stage 2 Results, Showing Standardized
Regression Coefficients
59
Figure 10. Path Analysis for Stage 3 Results, Showing Standardized 77
Regression Coefficients.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Abstract
This study examined the utility of the behavioral health services utilization model
to explain treatment entry for substance abuse problems among older males. Participants
in the study were 855 male veterans aged 55 and over, who were screened for alcohol
problems during inpatient medical treatment. Regression analysis found that treatment
seeking behavior was associated with age, marital status, alcohol problem severity, illegal
drug use, and cognitive status. Contrary to expectations, age was negatively associated
with treatment seeking. Age was mediated by alcohol problem severity. This result
indicated that severity only partially explained the relationship of age to treatment
seeking. The findings of this study may facilitate effective recruitment of older adults into
intervention programs, by identifying which problem drinkers are most ready to respond
to offers of assistance, and by providing an understanding of the factors that contribute to
readiness to enter treatment.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1
Background
This study aimed to impro ve the understanding of what influences older adults to
accept or refuse alcohol treatment. The study examined a model of treatment seeking
adapted to older male problem drinkers with multiple social and psychological
problems-a segment of the elderly population in particular need of substance abuse
intervention. The model of treatment entry examined in the current study was based on
the behavioral health services utilization model of Andersen (Andersen & Newman,
1973), which identified predisposing, need and facilitating characteristics in treatment
seeking. While researchers have begun to identify individual factors associated with
treatment seeking behavior, much of this work has focused on younger adults (Hingson,
Scotch, Day & Culbert, 1980; Kilpatrick et al., 1978; Pfeifer, Feuerlein & Brenk-Schulte,
1991; Weisner, 1993). Recent studies have also examined treatment seeking among older
adults (Brennan & Moos, 1991, 1996; Gomberg, 1995). These previous studies have
omitted components that may help to explain treatment entry in the elderly, such as
cognitive status, emotional distress, and comorbid drug use. The importance of these
factors is described in further detail below.
The study also sought to clarify the relationships between predictive factors,
which prior studies have not explored. Prior studies of younger adults have found a
positive association between greater age and treatment entry (Hingson et al). It was
hypothesized that if age was also associated with treatment entry in an older sample, it
could be influenced by several factors. This relationship might be influenced by alcohol
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
problem severity, which has also been associated with treatment entry (Gomberg, 1995;
Weisner, 1993). Social support and education are also mediators that may help to explain
treatment entry, yet have not been examined in samples of older adults. Comorbid drug
use may also influence treatment seeking in a population with multiple substance abuse
problems. In order to clarify these relationships, the study aimed to test a mediational
model in which the association of age to treatment entry was mediated by alcohol
problem severity, drug use, education, and social support, (see Figure 1). It was
hypothesized that each of these factors would contribute to a modified version of
Andersen’s behavioral model. A second model, which expanded the behavioral model by
incorporating the effect of distress and cognitive status, was also examined (Figure 2).
These models have the potential to explain the relationship of patient age to treatment
entry in an especially vulnerable segment of the older adult population.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Education
3
Figure 1. The behavioral model.
" O
v _______________________ _
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
D rug Use
4
Figure 2. The proposed developmental distress model.
• c
;
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
D ru g Use
5
Impact of alcohol problems on older adults. Among community samples, current
prevalence of DSM-IV alcohol abuse and dependence has been estimated at 1% to 3% for
men over age 65 and .5% to 1% for women over age 65 (Bucholz, Sheline & Helzer,
1995). Due to the medical and psychiatric pathology associated with heavy alcohol use,
clinical samples have consistently shown higher rates of prevalence than community
samples. This has been apparent from samples at VA hospitals, emergency rooms and
surgical facilities (Moos, Mertens, & Brennan, 1993; Schuckit, Atkinson, Miller, &
Berman, 1980; Tabisz et al., 1991). Prevalence of alcohol problems among adults over 60
in medical settings has been estimated at 20% (Curtis, Geller, Stokes, Levine, & Moore,
1989). In VA hospitals specifically, Gomberg (1975) found that 15% of male inpatients
over 60 were current drinkers with a diagnosis of alcohol dependence. Clearly older
problem drinkers appear in hospital settings in significant numbers, providing an
opportunity for targeted intervention.
Alcohol abuse and dependence among older people have considerable personal,
financial and social costs, both to those directly affected and to society at large. Physical
and mental health problems associated with excessive alcohol use are many.
Physiological systems negatively affected by alcohol include gastrointestinal,
cardiovascular, endocrine, hematological and neurological (see Gambert & Katsoyannis,
1995, for a review). Illnesses involving these systems are particularly prevalent and
serious among older adults. Excessive alcohol use in late life is also associated with
increased risk of cognitive impairment (Vogel-Sprott & Barrett, 1984). Other
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
consequences of alcohol dependence include social isolation, and increased risk of
anxiety, depression, and suicide (Blow, 1998). Despite these numerous problems,
treatment entry rates have been estimated at only 10% to 30% of medical patients with an
identified alcohol problem (Stephan, Swindle & Moos, 1992). As a result, there is a
pressing need to better understand the process of how older adults begin treatment for
alcohol problems.
Screening for alcohol problems. This study examined those factors that have been
reported to predict treatment entry, utilizing a medical patient sample of older adults who
had been identified as having a possible alcohol problem. In order to identify and treat
these individuals, a case finding approach has been suggested (Gomberg, 1995). This
entails the active screening and recruitment of individuals considered unlikely to seek
services on their own. Case finding for older adults has often focused on at-risk
populations, including medical patients and other clinical populations that have high rates
of alcohol problems (Moos et al. 1993). As described below, the study examined
treatment entry in the context of a case-finding approach to intervention.
Unfortunately, identification of individuals with alcohol problems in medical
settings does not imply successful entry into treatment. Of those individuals who are
successfully identified and offered treatment, a high percentage declines assistance, and
perhaps denies the existence of a problem (White, Luckie & Miller, 1996). In one study
of middle aged and older men within the Veterans Affairs medical system, only 10% of
individuals detected as having an alcohol problem eventually entered treatment (Stephan
et al., 1992). This finding is in agreement with general population studies by the National
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Institute on Alcohol Abuse and Alcoholism, which have estimated that only 10% of
alcoholics or problem drinkers ever receive treatment (U.S. National Institute on Alcohol
Abuse and Alcoholism, 1979). Clearly the vast majority of those with an alcohol problem
do not receive treatment.
This low rate of entry may have a significant negative impact on health and well
being, if in fact those older adults who refuse treatment continue their problematic
drinking behavior. Some older adults may reduce the quantity of alcohol they consume,
in accordance with concerns about their health (Dunham, 1981). However, few studies
have been conducted of spontaneous remission of alcohol problems in the absence of
treatment. Those that have examined this question have found mixed support for recovery
in the absence of formal treatment or “natural recovery,” with almost no attention to
natural recovery among older adults (see Gomberg, 1995, for a review). If in fact
successful entry into treatment is a prerequisite to cessation of harmful drinking, it
becomes even more important to understand the process by which some people enter
treatment, while others either refuse treatment outright or for some other reason do not
obtain the treatment that they need. The present study aimed to accomplish this by
examining a model of treatment entry in a sample of older male medical patients.
Models of Treatment Entry
The mediational model proposed in the current study was based on previously
tested theoretical explanations of health service usage. Several theoretical models have
been proposed to help explain why some people seek out health services and others do
not. These models have also been applied, with some modification, to treatment seeking
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
behavior for substance abuse and other mental health problems. In the section below, I
discuss the most promising of these models, the behavioral model. After reviewing the
support for this model, I propose a “developmental distress model” that draws elements
from the behavioral framework. This model incorporates the basic elements of the
behavioral model, and includes the additional components of psychological distress and
cognitive status. This model was examined as a promising explanation for observed
patterns of alcohol problem treatment entry among older adults.
The behavioral model. Andersen’s behavioral model has played an important role
in understanding treatment seeking behaviors for a number of different health problems
(Krause, 1990). According to Andersen’s behavioral model of treatment seeking, there
are three determinants of health service utilization: predisposing, facilitating and need
characteristics (Andersen & Newman, 1973). Predisposing characteristics such as
demographic factors exist prior to current illness onset, and influence individuals’
treatment seeking behavior. Facilitating factors include income, social support and other
factors that positively affect access to services. Need characteristics include both
perception of illness severity and objective measures indicating the presence of a
disorder. These three components are hypothesized to exert unequal influence on service
utilization. For physician visits, Andersen estimated the influence of illness severity to be
high, facilitating resources such as family support and demographic factors to be of less
importance (Andersen & Newman, 1973).
Studies examining the predictive utility of the behavioral model in substance
abuse treatment entry have found mixed results. Each of the three types of factors
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
included in the behavioral model have been found to have some association with
treatment seeking, although results are not consistent across studies. In a large
community study of drinking behavior, Hingson and associates (1980) conducted a
survey of 5320 adults in the Boston area. Of this sample, 2% indicated that they had a
current drinking problem. Of these, only 15% were currently receiving some form of
treatment or assistance for their drinking. Hingson and his associates found that among
those who self-identified as having a current alcohol problem, those who had sought
treatment contained a higher percentage of adults over 50 than the group who had not
sought treatment, a finding that lent support to greater age being a predisposing factor in
treatment seeking. Unfortunately, age range and distribution were not reported. Those
who claimed some religious affiliation or were better educated were also significantly
more likely to have sought help. Those who had stopped drinking in the past year were
also more likely to have sought formal treatment. No differences were found by race,
marital status, employment status or sex. These findings provided mixed support for the
importance of the predisposing factors identified by the behavioral model. In particular,
age was the most important demographic factor associated with treatment entry.
Unfortunately, the study did not examine the significance of other predictors in adults
over 50 years of age, as compared to those under 50, so conclusions regarding age group
related differences could not be drawn.
Like Hingson and his associates, Brennan and Moos (1991) conducted a
retrospective study that examined factors from the behavioral model in relation to self-
reported history of treatment. This study utilized a sample of 1315 older male veterans
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
who were drinkers. Treatment seeking was measured by asking participants if they had
sought help in the past year. They found that marital status was associated with treatment
seeking, with married people less likely to seek treatment. They also found that
individuals who consumed more alcohol and had more problems related to drinking were
more likely to seek treatment, although alcohol-related problems were better predictors of
treatment seeking than quantity of alcohol consumed. As with other studies, the
investigators found that late onset was associated with less severity (Schonfeld & Dupree,
1991). Consistent with lower alcohol dependence severity, older adults with recent onset
of alcohol problems were less likely to have sought treatment than those with early onset.
Early onset, on the other hand, was associated with greater severity and greater likelihood
of treatment seeking. This association of early onset, higher severity, and greater
likelihood of seeking treatment suggested that having suffered for a long period of time
with the effects of alcohol dependence, older people’s motivation for treatment may
increase in late life.
Weisner (1993) compared problem drinkers in treatment with those in the general
population. This study used intake data from several public alcohol programs, as well as a
general population survey of households in Northern California. The sample included 256
men in public treatment programs, and 134 untreated men in the general population.
Comparing these two groups in a chi-square analysis, Weisner found that those in
treatment were more likely to be African American, divorced, unemployed, and low
income. It is possible that socioeconomic variation between the two samples could have
accounted for some of these findings, since the treatment sample was drawn from public
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
programs. In a discriminant function analysis, these factors did not significantly
distinguish those in treatment from those not in treatment. Rather, being older and having
a history of prior treatment predicted enrollment in current treatment. These factors are
considered predisposing according to the Andersen model. Weisner also found that a
higher proportion of adults in treatment for alcohol problems reported a history of daily
drinking, scored higher on measures of alcohol-related social consequences such as
family, legal, job and health problems, and also scored higher on measures of alcohol
dependence severity. Weisner used discriminant function analysis to determine the
relative contribution of predisposing, need, and facilitating variables to a model
predicting treatment utilization. She found that need variables contributed a greater
proportion than predisposing and facilitating variables. These findings supported
Andersen’s hypothesis that need is of primary importance in predicting treatment entry.
In a study that examined treatment entry in older adults, including analysis of
some elements of the behavioral model, Gomberg (1995) compared two groups of older
adults with an identified alcohol problem-104 currently enrolled in treatment and 64 not
in treatment. Those in treatment scored significantly higher on measures of dependence
severity. Those in treatment also reported more alcohol-related negative consequences
such as complaints from children, excessive time spent drinking and reduced activities.
Those currently in treatment were also more likely to have received prior alcohol
treatment, which could be considered an indicator of severity as well as a predisposing
factor. No differences were found by marital status. While this study did not aim to test
the utility of the behavioral model as a whole, it provided additional support for the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
importance of need characteristics in predicting treatment entry for older adults.
The results described above provided support for the predictors identified by the
behavioral model. However, the methodology of these studies limits the conclusions that
may be drawn from them in regard to prediction of future treatment entry. The studies by
Hingson (1980), and Brennan and Moos (1991) relied on retrospective self-reporting of
treatment history. The studies by Weisner (1993) and Gomberg (1995) examined factors
that differentiated individuals in treatment from those not in treatment. It is not clear from
any of these studies how long the participants had been in treatment at the time data were
collected. Therefore, those factors that differentiated participants in treatment from those
not in treatment may be more important for treatment retention than for treatment entry—
the studies say nothing about what may have distinguished the two groups at the
particular point in time at which they had the opportunity to begin treatment.
In addition, the sampling methods of the studies by Weisner (1993) and Gomberg
(1995) were problematic. These studies compared a clinical sample to a community
sample, and then tried to draw conclusions about factors contributing to alcohol problem
treatment entry based on the differences observed between the two samples. In fact,
sampling inconsistencies may have accounted for the observed differences. A more
appropriate method, which the current study utilized, is a longitudinal prospective study
that examines possible predictors in a sample determined to have high need for treatment,
prior to actual treatment entry, and then examines subsequent treatment seeking behavior.
This methodology eliminates the potential sampling inconsistencies of previous studies,
and can test predictor variables that are collected prior to treatment entry.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Physical health. One aspect of treatment need that might be expected to play a
role in treatment entry for older adults is physical health. It has been hypothesized that
negative health consequences may motivate older adults to reduce drinking (Stall, 1987).
In a community survey of older adults, Dunham (1981) found that health problems led
40% of respondents to either quit or reduce drinking. In addition, treatment seeking
models for disorders other than alcohol problems have identified physical illness
measures as more important in predicting health service usage than beliefs about
treatment efficacy or other factors (Krause, 1990). However, studies have not found a
relationship between severity of physical health problems and alcohol treatment seeking
in older adults— the association between illness and reduction in drinking does not result
in greater enrollment for the physically ill in formal treatment programs. In fact,
Gomberg (1995) found that 24% of those in treatment and 61% of those not in treatment
reported hospitalization in the past year, a difference significant at the g < .001 level. This
seemed to indicate that health problems may form a barrier to treatment entry. Brennan
and Moos (1991) found no relationship between poor physical health (as measured by
number of physical health problems on a self-report inventory) and alcohol problem
treatment seeking in older adults. These findings suggested that while health problems
may motivate reductions in drinking in the elderly, physical illness may impede access to
formal treatment. If in fact physically ill older adults reduce their level of alcohol intake,
these findings suggest that they may accomplish this reduction without treatment.
Social support. Studies have generally found that a lack of social support is
associated with negative mental and physical health outcomes (Krause, 1987). Therefore,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
it might be expected that the presence of social support would function as a facilitating
factor in treatment seeking. Although it makes sense intuitively that social support would
facilitate effective intervention, a positive relationship of social networks or social
support to treatment seeking has not been clearly established. According to the behavioral
model, social support functions as a facilitating factor that makes treatment entry easier.
Finlayson, Hurt, Davis and Morse (1988) found that older people most often cite the
encouragement of family to get help as a motivating force in treatment seeking.
Therefore, it seemed likely that married men would have additional encouragement to
seek treatment, as well as possible practical support that would make such treatment
possible. Yet studies examining support have found no evidence for such facilitation with
regard to alcohol problems. The studies by Weisner (1993) and Brennan and Moos (1991)
described above found that men in treatment were less likely to be married than men not
in treatment. Studies by Gomberg (1995) and Hingson et al. (1980) found no relationship
between marital status and treatment. Brennan and Moos also measured the amount of
social support received by study participants, and found that older problem drinkers with
more family and friend resources were less likely to seek treatment. The investigators
hypothesized that family may have actively discouraged older drinkers from seeking
treatment, possibly due to the social stigma attached to alcohol problems. Family
members who drink might also have discouraged treatment entry. These equivocal
findings indicate that the role of social support as a facilitating factor in treatment seeking
needs further investigation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In summary, studies that have tested elements of the behavioral model have found
mixed support for its value in predicting who seeks treatment, as well as inconsistent
support for the predictive value of the individual elements of the model. Those predictors
with the most consistent empirical support have included the predisposing factor of age,
and need characteristics such as alcohol symptom severity. Of the three categories of
predictors in the model-predisposing, need, and facilitating factors, need appeared to
have the most consistent empirical support.
Although these findings support some aspects of the behavioral model, the
explanation for treatment entry offered by this theory could also bear further elaboration
and development. Regarding the relationship of greater age to treatment entry, it was not
clear whether this association would be found in an elderly sample. The studies described
above in which this finding occurred included younger and middle-aged samples. The
present study examined this factor within an elderly sample. Using this different age
range, the current study aimed to determine if there was also a correlation of greater age
to higher rates of treatment entry in late life.
It was also proposed that the behavioral model might need greater elaboration. In
a review of the utility of the behavioral model, Krause (1990) pointed out that while
studies supporting the behavioral model have found that there are significant differences
between those who seek treatment and those who do not, the model as a whole does not
generally explain more than 20% of the variance in treatment seeking. Therefore, the
model may benefit from incorporation of other elements in order to better explain
treatment seeking. These proposed modifications to the model are discussed below.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
16
Adapting Treatment Entry Models to Older Drinkers in a Medical Setting
The behavioral model reviewed above has shown promise in helping to explain
why some people enter treatment while others refuse. However, this model does not
include several factors that may be important in understanding why older adults seek
treatment. These include mental health concerns of particular importance to elderly
problem drinkers, such as anxiety, depression, and cognitive status. A second model was
proposed to explain the relationship of these factors to each other and to treatment entry
(Figure 2). The importance of these factors in late life, and their potential impact on
models of treatment entry are discussed below.
Emotional distress. Prior theoretical models and empirical studies have paid little
attention to the role of emotional distress in help-seeking behavior for alcohol problems.
However, it was hypothesized that distress might function as a source of additional
motivation for treatment entry. In an explanatory model of treatment entry, distress may
mediate the relationship of severity to treatment entry (See Figure 2). Numerous studies
have found high rates of comorbid anxiety and depression with alcohol abuse. It has been
hypothesized that alcohol is used to relieve symptoms of depression and anxiety, but
when used to excess actually increases symptoms of these disorders in the long run
(Barlow, 1988; Mueller et al., 1994). Studies of patients in treatment have found that
depression and anxiety are correlated with alcohol problem severity (Velasquez,
Carbonari & DiClemente, 1999). These findings suggest that the relationship of alcohol
problem severity to treatment entry may be mediated by psychological distress, or vice
versa. If individuals have an alcohol problem yet experience neither anxiety nor
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
17
depression, they may feel insufficient motivation to seek treatment.
Studies comparing comorbidity across age ranges suggest that rates of comorbid
anxiety and depression are even higher in older alcoholics than in younger ones, possibly
due to the long-term negative effects of alcohol dependence on psychological health. This
trend is the reverse of prevalence studies based on community samples, which show
declines in psychopathology with age. In a large (N = 22,463) sample of male veterans,
age-related increases were observed for major depression, anxiety disorders, and
dementia (Blow, Cook, Booth, Falcon, & Friedman, 1992). Rates of comorbid substance
abuse disorders, posttraumatic stress disorder, schizophrenia, and personality disorders
peaked in younger and middle-aged patients and then declined with age. Comorbid
anxiety disorders have been estimated at 1% to 9% in clinical treatment samples (Blow et
al., 1992; Finlay son et al., 1988). Estimates of mood disorder occurring in older alcohol
abusers vary from 12 to 30% (Blow et al., 1992; Finlay son et al., 1988). If comorbid
depression and anxiety do increase with age, the role of these disorders seems especially
important to consider in a model of late life alcohol problem treatment seeking.
The relationship of depression to treatment seeking among older adults has not
been established, though some preliminary studies suggest that a link exists. Depression
is often associated with poor motivation, and hence might be expected to impede
treatment entry. However, there is some evidence from studies of younger adults that
depression may be positively associated with treatment entry or positive treatment results.
In a small study, Woodruff (1973) compared 25 adults in treatment for alcohol
dependence with a similar sample who did not seek treatment, and found that the group in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
treatment was more depressed. In a similar study, Chan, Pristach and Welte (1991)
compared adults in treatment for alcoholism to non-treatment groups with approximately
equal drinking habits and found that depression was higher in the treatment group.
Another study found that higher depression scores at the time of treatment were
positively associated with remission from drinking in adults over age 60, but not in
younger adults, suggesting that depression may have a cumulative effect of increasing
motivation over time (Helzer, Carey, & Miller, 1986). Brennan and Moos (1991) also
found that depression was associated with greater rates of alcohol treatment seeking in
older adults. These studies suggest that among older adults, depression in conjunction
with an alcohol problem may result in increased motivation to enter treatment for alcohol
problems.
The relationship of anxiety to treatment seeking has been studied even less than
depression. However, there is some indication that anxiety may also increase treatment
entry. In a study of middle-aged adults (mean age of 45), Kilpatrick and his associates
(1978) administered psychometric tests to 167 male veterans who were chronic alcoholics
enrolled in inpatient treatment. They found that participants who rated as highly
motivated for treatment had higher scores on the MMPI psychasthenia scale, which
measures trait anxiety, scored higher on a mood state measure of tension/anxiety
according to the Profile of Mood States (POMS), and also scored higher on both state and
trait subscales of the Spielberger State-Trait Anxiety Inventory, as compared to adults
who were rated as not motivated for treatment. These findings suggested that a high level
of anxiety may be an important predictor of treatment entry, possibly providing additional
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19
motivation for individuals to resolve their drinking problem. Since these findings seemed
to indicate a high level of distress associated with alcohol problems, anxiety might
therefore be conceptualized as need characteristic. If older adults with an alcohol problem
are likely to be even more anxious than younger adults, the role of anxiety in predicting
treatment entry is an important area of investigation.
Comorbid use of drugs and alcohol. Comorbid drug use is also an important
consideration among indigent older men. While use of illegal substances such as heroin
and cocaine is not prevalent in the general older adult population, it is a significant
problem among the population from which the present sample was drawn. Potential
adverse interaction of either prescription drugs or street drugs with alcohol poses
particular risk for older adults (Blow, 1998). However, the potential role of drug use in
treatment seeking has never been explored in older populations. In the current study, it
was hypothesized that comorbid drug use would function as an indicator of substance
abuse severity, and would therefore increase the likelihood of treatment seeking.
Cognitive status. Of particular concern for older drinkers, cognitive functioning
has not been examined as a predictor of successful treatment entry. Yet this may be an
important indicator of the ability of an individual to participate meaningfully in treatment.
Without adequate cognitive functioning, an older person may be unable to make use of an
outpatient treatment program. Even if participation was feasible, even a mild level of
impairment might make logistical aspects of treatment entry, such as remembering how
to get to therapy appointments, too difficult. Therefore, cognitive status may be
conceptualized as a facilitating factor within the behavioral model, with higher cognitive
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
20
functioning positively associated with successful treatment entry.
Screening for cognitive status may be particularly important in samples of older
male veterans, in which prevalence of dementia is high. One study using a large sample
of veterans with a presenting diagnosis of alcohol dependence who had sought outpatient
psychiatric treatment found comorbid dementia or organic brain syndrome in 9% of those
60 to 69 and 18% of those over 70 years old (Blow et al, 1992). These percentages are
higher in comparison to prevalence rates in clinical samples of older adults, which have
also found that rates of dementia increase with age from approximately 5% among older
adults in their sixties to 12% among adults over 70 (Ritchie, Kildea, & Robine, 1992).
This elevated prevalence is possibly attributable to the direct effect of heavy alcohol
consumption on cognitive functioning in those older adults with a long-term alcohol
problem (Tarter, 1995; Vogel-Sprott & Barrett, 1984). Whether impairment is due to
excessive alcohol consumption or to other causes, clearly cognitive status is likely to be
an important facilitating factor in treatment entry if some individuals are too impaired to
participate meaningfully in standard treatment programs.
Relationship between factors within the model. The literature on the behavioral
model reviewed above has identified factors that are likely to predict treatment entry for
older adults. However, studies conducted to date have not offered an adequate
explanation for the relationship between the constituent factors in the behavioral model: it
is not clear how predisposing, need and facilitating characteristics interact in predicting
treatment entry. Relationships between these factors are discussed in the section below
(see also Figure 2).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
21
Of particular importance to treatment entry in late life, there is some indication
that older adults are more likely to enter treatment than younger adults (Hingson et al.,
1980; Weisner, 1993). This hypothesis is consistent with developmental conceptions of
alcoholism that posit that alcohol problems progress though stages of increasing severity,
often over a course of decades (Gomberg, 1980; Jellinek, 1960). This developmental
conception is consistent with studies of early onset versus late onset alcohol problems in
older adults, which have found that early onset problems are more severe (Atkinson,
Tolson, & Turner, 1990; Brennan & Moos, 1991). It is also consistent with findings of
studies that have found that older adults with a late onset problem are less likely to seek
treatment than those with an early onset problem (Blow, 1998). These findings suggest
that the relationship between age and treatment entry is mediated by severity of an
individual’s alcohol problem.
It was also hypothesized that the relationship of alcohol problem severity to
treatment seeking could be mediated by anxiety and depression. Greater alcohol problem
severity has been associated with higher levels of anxiety and depression (Velasquez et
al., 1999). As discussed above, the few studies conducted to date on the relationship of
psychological distress to alcohol treatment seeking suggest that the two factors are
associated (Brennan & Moos, 1991; Kilpatrick et al., 1973). In the absence of
psychological distress, it seems unlikely that an individual would be sufficiently
motivated to seek treatment. Therefore, it was expected that psychological distress would
mediate the relationship of alcohol problem severity to treatment entry.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
22
It was expected that cognitive functioning would act as a mediating variable in the
proposed model. With increasing age, the prevalence of cognitive impairment increases
(Ritchie, Kildea & Robine, 1992). Cognitive impairment is also associated with greater
alcohol problem severity and longer duration of alcohol problems (Blow, 1998; Vogel-
Sprott & Barrett, 1984). Sufficient cognitive capacity may be necessary for effective
participation in treatment, and may also be an important facilitating factor in treatment
entry. Therefore, it was expected that cognitive impairment would mediate the
relationship of age to treatment entry as well as the relationship of alcohol problem
severity to treatment entry.
Studies have found that married men may be less likely to seek treatment than
unmarried men (Brennan & Moos, 1991; Weisner, 1993). However, this relationship may
be mediated by alcohol problem severity and psychological distress. For men with an
alcohol problem, being single in late life may be indicative of a longstanding alcohol
problem that makes a marital relationship difficult or impossible. Alternatively, the stress
associated with divorce or widowhood may lead to greater levels of distress and alcohol
problem severity among men (Kasl, Ostfeld, Berkman, & Jacobs, 1987). These
hypotheses are supported by studies that have found that early onset elderly alcoholics are
more likely to be divorced or widowed than those with a late onset problem (Schuckit &
Pastor, 1979). Studies have also found that being divorced or widowed is associated with
depression in late life (Krause, 1987). Based on these studies, it was expected that
married individuals would indicate reduced levels of psychological distress compared
with those who were not married, as well as alcohol problems of lower severity.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
23
Therefore, it was expected that the effect of being single on treatment entry would be
mediated by alcohol problem severity and psychological distress.
Summary
Andersen’s behavioral model has identified several factors associated with
treatment entry. However, a satisfactory theory explaining treatment entry has yet to be
developed with regard to older adults. The model examined in this study is a modified
version of Andersen’s behavioral model, incorporating factors of particular importance to
treatment entry for older adults with an alcohol problem. The predictors of treatment
entry in this model included predisposing characteristics (age, education and marital
status), need characteristics (alcohol problem severity), facilitating factors (cognitive
status) comorbid psychological distress (anxiety and depression). According to the
developmental distress model, it was expected that the male veteran most likely to seek
treatment would be older, not married, with a longstanding, severe alcohol problem, who
has symptoms of comorbid depression and anxiety, and has comparatively higher
cognitive functioning.
This study examined data gathered in the course of a screening and treatment
program to test the predictive value of the proposed models. It was expected that this
information would be potentially very important in improving low rates of treatment
entry, helping clinicians to target who is likely to respond to offers of treatment (thus
saving precious time and resources) or alternatively to refine screening and intervention
methods in a way that will help to build motivation to seek help (Prochaska et al., 1992).
This is an important factor in treatment planning, because interventions tailored to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
24
individuals with no intention to quit drinking should be fundamentally different from
those targeting individuals who are ready to change their behavior. Without appropriate
matching of interventions and participants, treatment is less likely to be effective (Miller
et al, 1995).
A particular strength of the study was the utilization of measures to predict
treatment entry that were administered in the course of treatment screening rather than
after treatment had begun. The study was the first known of this type. This enabled the
identification of those factors that predict treatment entry rather than treatment retention.
While both treatment entry and treatment retention are important outcomes, the literature
reviewed above has frequently studied the two in a way that does not make their
differences clear. Many of the studies testing the utility of the behavioral model in
alcohol treatment entry do not take this distinction into account (e.g., Chan et al., 1991;
Gomberg, 1995; Weisner, 1993; Woodruff, 1973). Therefore, these studies have not been
able to distinguish predictors of treatment entry from predictors of treatment retention.
The current study aimed to specifically examine factors related to treatment entry by
analyzing data concerning possible predictors that had been gathered prior to actual
treatment entry. A study that examines predictors of treatment entry for older adults using
data gathered prior to actually beginning treatment has yet to be conducted, and will
make a significant contribution to understanding treatment entry in late life.
Hypotheses
In accordance with the models proposed above, it was hypothesized that the following
factors would be associated with treatment entry:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
25
1. Predisposing factors: Age (older), education (higher), and marital status (not married).
2. Facilitating factors: Cognitive status (higher).
3. Need characteristics: Alcohol symptom severity (higher) and drug use (present).
4. Emotional distress: anxiety and depression (higher).
In addition to testing the independent effects of these independent variables, the
following hypotheses were tested regarding the relationship of the different factors in the
model:
1. The relationship of age to treatment seeking will be mediated by alcohol problem
severity, drug use, and education.
2. The relationship of alcohol problem severity to treatment entry will be mediated by
depression and anxiety.
3. The relationship of alcohol problem severity to treatment entry will be mediated by the
effect of cognitive impairment.
4. The relationship of marital status to treatment entry will be mediated by the effects of
severity, depression and anxiety.
Method
Participants. Between 1991 and 1999, 1366 male veterans over the age of 55 at
the West Los Angeles VA Hospital were identified by medical staff as having possible
problems either with substance abuse or misuse of their medications (including non
psychotropic medications). These men were referred to the GET SMART program, a
screening and treatment program designed specifically for older adults in the VA system,
for further evaluation and for possible entry into group psychotherapy. This program
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26
served a low-income population with high psychiatric comorbidity, approximately 30 to
40% of which were homeless or marginally housed in shelters or residential hotels.
Records completed by these individuals in the course of screening and evaluation
procedures for the GET SMART program were examined in the current study. Approval
for the use of these records was obtained from the West Los Angeles VA Hospital
Institutional Review Board (IRB). Approval forms are attached as Appendix A.
Screening procedures. 1366 participants were interviewed using a two-page
screening instrument designed to identify substance abuse and medication problems, and
to give medical staff a preliminary indication of severity of substance abuse problems.
This brief questionnaire included demographics (age, race, marital status, years
education), questions regarding frequency of drinking and drug use, the CAGE scale, and
indicated whether or not the participant was interested in receiving substance abuse
treatment. Of this sample, 855 received a score of one or higher on the CAGE, indicating
a possible alcohol problem. The others received a score of zero, suggesting no history of
an alcohol problem based on self-report. All individuals who were screened were offered
treatment, regardless of their score on the CAGE. About 253 individuals indicated that
they were interested in receiving treatment at the time of screening.
Approximately 250 individuals attended appointments with clinical staff for
further evaluation prior to entering treatment. Evaluation procedures included some
individuals who had initially said that they were not interested in treatment, and then
agreed to attend, as well as individuals who had incomplete initial screening information.
124 individuals began treatment. See Figure 3.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 3. Destination of individuals screened for treatment.
Medical Patients Scored 1 or Higher on the CAGE.
(N = 855)
Patients asked: “Are you interested in treatment?’
1 i
Who began
treatment?
I
Said “Yes” Said “No” Missing data
(n = 253) (n = 500) (n = 102)
Who showed up
for evaluation?
Yes
(n = 250)
No
(n = 605)
Yes No
(n= 124) (n = 731)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
28
Clinical staff consisted of a geriatric psychiatrist, a registered nurse with a Ph.D.
in psychology, and a clinical psychology intern, each of whom received extensive
training in the assessment measures used. Instruments administered included the MAST-
G (Blow et al., 1992) a 24-item self-report measure of substance abuse problem severity
designed specifically for older adults, the Hamilton Anxiety Scale (Hamilton, 1959), the
Hamilton Depression Scale (Hamilton, 1960), the Mini Mental Status Exam (Folstein,
Folstein, & McHugh, 1975), and a drinking history questionnaire that included age of
onset and past treatment history. All individuals who completed this further screening
process were offered treatment in either an unstructured support group or a cognitive
behavioral treatment group (group assignment was not random). About 50% began
treatment, while the remainder dropped out or lost contact with hospital staff, and did not
enter treatment at that time.
After the initial screening, the evaluation and treatment record collecting
procedures were complicated by a number of factors. The sample examined in this study
was a difficult one to track, including a large percentage of individuals with multiple
substance abuse problems, psychiatric problems, and homelessness. Of particular
concern, individuals often completed only part of the evaluation process, and then lost
contact with hospital staff. Changes in the VA hospital staff and inconsistent procedures
made the records gathered in the course of evaluation incomplete. After the point of
initial screening, it became somewhat more difficult to follow participants for the purpose
of examining predictors of retention treatment entry. These problems made the second
proposed model, which incorporated distress and cognitive functioning, difficult to test
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
29
because of small sample size and resulting low statistical power. In accordance with these
limitations, results based on the evaluation stage of treatment entry should be considered
preliminary.
Instruments used in Screening
Screening fonn. The form used initially to screen participants (n=1366) for the
study was two pages long. It contained demographic information (age, sex, race, religion,
marital status, years of education), the CAGE questionnaire (Ewing, 1984), and questions
regarding presence of drug use. The CAGE questionnaire is a widely used, 4-item
screening instrument, which has been validated for use with many different populations,
including older adults (Buchsbaum, Buchanan, Welsh, Centor, and Schnoll, 1992). In
older male samples, the CAGE has a sensitivity of 86% and specificity of 78% for a cut
off score of one (Buchsbaum et al., 1992).
Instruments Used for Further Evaluation
Drinking history inventory. This form was completed by participants at the
beginning of treatment. This form asks participants when they initially began using
alcohol and/or drugs, what type of drugs or alcohol they used, which emotional or
environmental triggers to use drugs or alcohol they experienced, whether they are
currently drinking, and how extensive their treatment history was. Questions about
treatment history addressed past inpatient and outpatient treatment, and attendance at
Alcoholics Anonymous. Studies have suggested that self-report measures of alcohol
consumption among older adults are basically reliable, in that they are consistent with
collateral reports (Tucker, Vuchinich, Harris, Gavomik, & Rudd, 1991). Data obtained
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
30
from this form includes duration of drinking problem as well as treatment history
information. Because most of the individuals who completed this form began treatment,
responses were not considered useful for use in predicting treatment entry. However,
responses were examined in relation to drug and alcohol severity measures. The Drinking
History Inventory was administered to 148 individuals.
MAST-G. The Michigan Alcohol Screening Test-Geriatric Version (MAST-G,
Blow et. al., 1992) is a 24-item instrument designed to detect alcohol abuse among older
adults. It is a self-report measure of symptoms, asking respondents to answer true or false
to each symptom listed. Five factors underlying the scale include alcohol dependence,
loss of control with drinking, loss and loneliness, rule-making and relaxation. The
MAST-G has a sensitivity of 94% and specificity of 78% in detecting alcohol dependence
in older adults, based on analysis of the instrument in a sample of 305 older adults with
widely varying levels of drinking, and has also shown adequate reliability (Blow, Brower
et al., 1992). In the current study, results of reliability analysis indicated that the measure
had adequate reliability, a = .89. The MAST-G was used as an indicator of alcohol
problem severity, and was completed by 160 individuals.
HRS-A. The HRS-A and the HRS-D were used as measures of psychological
distress for the purpose of the analyses. The Hamilton Rating Scale for Anxiety (HRS-A,
Hamilton, 1959) was designed to measure the severity of anxiety in patients already
diagnosed as suffering from an anxiety disorder. It contains 14 items measuring
psychological and somatic components of anxiety. Each item is rated by the interviewer
on 5 levels of severity from none (0) to very severe (4), based on the presence of specific
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
31
symptoms. In a study that examined the use of the HRS-A with an elderly sample, the
scale showed good discriminant validity in the identification of individuals with
generalized anxiety disorder (94% correctly identified), and strong inter-rater reliability r
= .82) (Beck, Stanley, & Zebb, 1999). The HRS-A has been found to have reliability of
.78 when used with younger adults (Cronbach’s alpha) (Riskind, Beck, Brown, & Steer,
1987). In the current study, results of reliability analysis indicated reliability consistent
with these prior findings, a = .78. The HRS-A was administered to 84 individuals.
HRS-D. The Hamilton Rating Scale for Depression (HRS-D, Hamilton, 1960) is a
widely used clinical rating scale for assessing severity of depression, consisting of 21
items that are each rated in terms of severity. Somatic symptoms, mood, suicidality and
apparent insight items are included, severity of which is rated by an interviewer, based on
the presence of specific symptoms. The HRS-D has had good inter-rater reliability when
used with older adults r = .92) (Beck et al., 1999). The HRS-D has also been found to
have strong discriminant validity in distinguishing anxiety from depression in younger
samples r = .78) (Moras, DiNardo, & Barlow, 1992). The HRS-D has reliability of .73
(Cronbach’s alpha) when used with younger adults (Riskind et al., 1987). In the current
study, results of reliability analysis indicated adequate reliability, a = .83. The HRS-D
was administered to 81 individuals.
MMSE. The MMSE was used as a measure of cognitive functioning. The Mini
Mental State Exam (MMSE, Folstein, Folstein, & McHugh, 1975) is a widely used
dementia screening instrument. This 30-point scale includes brief measures of orientation,
attention and concentration, memory, language and executive control. The MMSE was
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
32
used as a predi ctor of treatment entry for those participants who were not so impaired that
they were eliminated from the study. Reliability for the MMSE has been estimated at .89
(Folstein et al., 1975). Results of reliability analysis in the current study indicated
adequate reliability, a = .73. The MMSE was administered to 90 individuals.
Results
Overview of Analyses
Data were analyzed in three stages. At the first stage of the analysis, those
hypotheses testable using the screening form alone were examined (the first model,
Figure 1). These analyses included all participants who completed the initial screening
form and who received a score of one or more on the CAGE (n = 855). Descriptive
statistics were examined for all variables of interest in the study, including correlations
between demographic measures and CAGE scores. T-tests and chi-squares were used to
examine differences between those individuals who expressed interest in treatment and
those who indicated that they were not interested in treatment.
A regression analysis was conducted to examine the relative importance of
demographic measures versus substance abuse measures in predicting expressed interest
in treatment. This took the form of a hierarchical logistic regression model. Demographic
variables including age, education and marital status were entered in a regression model
first, followed by CAGE score and self-reported past use of illegal drugs. The dependent
variable for this analysis was the participant’s indication of interest in treatment.
Hypotheses regarding mediational effects of severity on age and marital status in
predicting treatment interest were also examined.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
33
At the second stage of the analysis, the first model was also tested (Figure 1). The
same hypotheses were tested using the same methods, except that the dependant variable
was the participants’ attendance at a treatment evaluation. In other words, this analysis
examined predictors of who showed up to be evaluated, rather than what individuals had
previously expressed about wanting treatment. This regression analysis examined the
utility of demographic factors and severity in predicting which participants attended the
evaluation. Mediational hypotheses were also tested.
At the third stage of the analysis, hypotheses regarding the independent effects of
each of the scales completed during the evaluation process were evaluated. In addition to
demographic and alcohol problem severity measures, this phase included descriptive
statistics for measures of depression, anxiety and mental status. Mean scores of those who
entered treatment and those who did not were compared on demographic, substance abuse
problem severity, and mental health measures. This included the effect of age, marital
status, alcohol problem severity as measured by the MAST-G, mental status and
psychological distress. The regression analysis at this stage incorporated results of these
measures in a regression analysis in order to test the second model (Figure 2). In this
analysis, the dependent variable was the individual’s presence or absence at the first week
of treatment.
Stage 1 Results
1366 men over the age of 55 were interviewed for possible participation in either
substance abuse treatment or a medication management program, and were given the
CAGE scale. 855 scored 1 or higher on the CAGE, suggesting a possible alcohol
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
34
problem. Of these 855, there were 753 whose screening forms indicated whether or not
they were interested in treatment. There were 102 who had incomplete screening forms,
which did not indicate whether the individuals were offered treatment. Missing responses
on this question were due to interruptions in the screening process and note taking errors
on the part of VA hospital staff conducting screening. Individuals with missing data on
this point did not differ from those without missing data in terms of age, race, marital
status, education, CAGE score or drug use. Analyses at this stage included the 753 men
who received a score of 1 or higher on the CAGE at the time of screening, indicating a
possible alcohol problem, and whose responses on the initial screening form indicated
that they were offered alcohol treatment.
Demographic characteristics. Participants ranged in age from 55 to 91, with a
mean age of 68.62 (sd = 7.02). Table 1 shows demographic characteristics of the sample.
See also Figure 4, which illustrates the age distribution of the sample. Age distribution
appeared to be slightly skewed in a positive direction, but skewness was not significant
(skewness = .61). Years of formal education ranged from 2 to 22 years, with a mean of
11.74 years (sd = 2.98), with education data missing in 3% of cases.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 1
Demographic Composition of the Stage 1 Sample (N=753)
Mean Age
Mean Years Education
68.62 (7.02)
11.74(2.98)
Ethnic Group
a
Percent
African American 303 40.2
White 367 48.7
Hispanic 61 8.1
Other 17 2.3
Missing 5 0.7
. Status n Percent
Married 240 31.9
Remarried 3 0.4
Widowed 104 13.8
Separated 23 3.1
Divorced 257 34.1
Never Married 108 14.3
Missing data 18 2.3
Figure 4. Age distribution of the sample at stage 1 (N = 855).
12 - : ------------------------------------------------------------------------------------------------------------------------------------------
c
C O
• ■ E
ra
Q .
ii
E
3
z
55.0 60.0 62.7 65.3 68.0 70.7 73.5 76.6 79.8 83.9
58.0 61.4 64.0 66.7 69.4 72.1 74.8 78.1 81.8 91.3
Age at Screening
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
36
The sample was 40.2% African American, 8.1% Hispanic, 48.7% White, and
3.4% other, with .5% missing. 32.3% of the sample was currently married, 3.1% were
separated, 34.1% were divorced, 13.8% were widowed, and 14.3% were never married.
Marital status data was missing for 2.2% of the sample.
Alcohol problem severity. As noted above, all individuals included in the stage 1
analysis scored 1 or higher on the CAGE, a screening instrument with a maximum score
of 4. Mean CAGE score was 2.29, with a standard deviation of 1.09.
Individuals were also asked how often they were currently drinking. Data on
current drinking were collected on 581 individuals (77.2%). Data on current drinking
were missing for 172 individuals (22.8%), due to omission of this step in early screening
procedures conducted from 1991 to 1994. Of those individuals with current drinking data
collected, 27.5 % said that they were drinking daily or almost daily, 7.4% a few times a
week 4.3% once a week, 6.9% one or two times per week, 13.9% once a month, and
39.9% said that they were not currently drinking. A one-way ANOVA was conducted to
examine the relationship of CAGE scores to frequency of drinking. Results found that
there were significant differences between the groups, F (575, 5) = 10.41, g < . 001. Mean
scores appeared to show a curvilinear relationship to drinking frequency and CAGE
score, with those drinking daily or not at all scoring the highest on the CAGE, and those
drinking once a week or once a month scoring the lowest. ANOVA contrast analysis
found the quadratic term to be significant, F (575,1) = 25.51, g < .001. These results
suggest that current abstention does not preclude past alcohol problems. See Figure 5.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
37
Figure 5. The relationship of mean CAGE score to drinking frequency (N = 581).
3.0
2 .8 -
2.6
2.4-
2.2 -
< D
u.
2,0 -
8
C /D
< u
O )
(0
O
■H
Drinking Frequency
It may seem counterintuitive to include those abstaining from alcohol
consumption in a sample of potential treatment candidates. However, as Figure 5
illustrates, self-reported abstainers had higher CAGE scores than those who reported
moderate drinking. This indicates that current abstainers were likely to have a history of
problem drinking. Such individuals could benefit from treatment, gaining assistance in
remaining abstinent. In addition, there is the possibility that individuals in the study
under-reported current drinking, despite a willingness to admit past alcohol problems.
Lastly, cessation from drinking may indicate a readiness to change a pattern of addictive
behavior, by making a step in the direction of sobriety (Prochaska, DiClemente, &
Norcross, 1992). For these reasons, abstainers were not eliminated from the sample.
Because abstainers reported more problems with alcohol than moderate drinkers (as
measured by the CAGE), frequency of drinking was not considered indicative of a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
38
drinking problem per se. For this reason, frequency of drinking was not included in the
proposed model as a predictor of treatment seeking behavior.
This sample included individuals with multiple substance abuse problems in
addition to alcohol abuse. At the time of screening, 28.0% of those who scored one or
higher on the CAGE also responded positively when asked if they had ever used other
drugs in addition to alcohol. Of those who admitted drug use other than alcohol, 47.2%
admitted using more than one substance. Most commonly used drugs included cocaine
(46.5%) and marijuana (40.2%), followed by heroin (36.0%), methadone (18.0%),
sedatives, tranquilizers or hypnotics (11.6%) and amphetamines (8.7%). 64.1% denied
using drugs. Data were missing on this question for 7.8% of the sample. In contrast to
clinical reports that have identified prescription psychotropic medications as of primary
concern in treating drug use other than alcohol among older adults (Blow, 1998),
individuals in the current sample favored use of illicit drugs.
Given the likely stigma attached to admitting illegal drug use to staff people in the
context of Veterans Hospital treatment, the percentage of self-reported drug use (28 .0%)
is probably an underestimation of drug use prevalence in the sample. The presence of past
drug use was associated with higher alcohol problem severity as measured by the CAGE.
Those who admitted past drug use had a mean score of 2.65 (1.10) on the CAGE, while
those who did not scored 2.10 (1.06), t = 6.19, p < .001.
Relationship of demographics to severity measures. One of the hypotheses of the
current study was that treatment seeking behaviors might be associated with age, due to
increasing severity of alcohol problems over time. This hypothesis was based on studies
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39
of younger adults showing an association of older age with treatment entry, and the
developmental theory that alcohol dependence is a progressive disorder that has a
tendency to worsen with duration. Even though population studies have found that older
adults drink less than younger ones, it was hypothesized that in a sample of problem
drinkers, severity might worsen with age. However, in this sample, older age was
associated with lower levels of severity as measured by the CAGE, r = -.26, g <.001. See
Figure 6.
Figure 6. Mean age of participants by CAGE score (N = 855).
7 2 ------------------------------------------------------------— -------------------— ----------- ---------------------------------------------------------------------------------------------------------
71-
70-
69-
6 8 -
o >
c
c
< 0
2
67-
o
C O
6 6 -
(0
0 )
03
<
Cage Score
Those who admitted illegal drug use were significantly younger, with a mean age of
65.68 (5.58) than those who had not used drugs, who had a mean age of 69.99 (7.13), t =
7.80, g <.001. A one-way ANOVA showed significant age differences based on
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
frequency of drinking, F (582, 5) = 4.10, g < .001. Mean scores showed a curvilinear
relationship between age and frequency of drinking. ANOVA contrast analysis found the
cubic term to be significant, F (585, 1) = 9.62, g = .002. Results suggested that older
individuals were more moderate in their drinking, with the oldest subjects drinking either
once a week or once a month, and the youngest subjects drinking daily, a few times per
week, or not at all. These results were consistent with those of population drinking
studies, which have shown a decline in drinking with age, rather than an increase in
drinking over time associated with prior alcohol abuse studies (Bucholz et al., 1995). See
Figure 7 regarding the relationship of age to drinking frequency at stage 1.
Figure 7. Age of participants by drinking frequency (N =581).
72 —
71 -
70-
69-
05
C
68-
0)
w
O
67-
C O
66-
0)
C D
<
6 5 ___
\
Drinking Frequency
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
41
In addition to lower alcohol problem severity, older age was also associated with
some demographic features that may have negatively influenced treatment seeking. In
particular, older age was significantly associated with lower levels of education r = -.17, p
<.001. This result was consistent with demographic studies that have found that earlier-
born cohorts have had fewer years of formal education than later-born cohorts (Schaie,
1996). Age was also associated with marital status. Those who were married had a mean
age of 70.02 (6.79), compared to those who were not married, who had a mean age of
67.89 (6.99), t = 3.92, p < .001.
Analysis of the effects of race focused on differences between African Americans
and Whites, since other ethnic groups were represented in small numbers. Race was not
associated with age, but was associated with differences in substance abuse severity,
marital status and education. African Americans scored significantly higher on the CAGE
(t = 2.88, p = .004). They were also more likely to report past drug use, X2 (1, N = 620) =
32.05, p < .001. African Americans were less likely to be currently married than Whites,
X2 (1, N = 656) = 6.01, p = .015, and reported significantly fewer years of education, t =
6.57, p < .001. See Table 2 regarding differences between African Americans and Whites
at stage 1.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
42
Table 2
Differences between African Americans and Whites at Stage 1 (N= 670)
African American fn= 3031 White fn= 3671
Mean Age 68.00 (6.98) 69.00 (6.96)
Years Education*** 11.09(2.91) 12.56 (2.72)
Percent Married* 26.6% 35.1%
CAGE Score** 2.42(1.08) 2.18(1.09)
Percent Drug Use*** 39.0% 20.2%
Want Treatment 37.7% 30.5%
*P < .05. **g < .01. ***p c .001.
Marital status was also associated with alcohol problem severity. Univariate
ANOVA found significant group differences in CAGE score by marital status, F (4, 817)
= 6.34, p < .001. Those who were married scored lowest, with a mean score of 2.05
(1.04), followed by those who had never married, 2.27 (1.12), were widowed, 2.30(1.17),
divorced, 2.41 (1.11), and separated, 2.93 (.83). Comparisons were also made examining
those who were married to those who were not, since the studies of health behavior have
largely examined the role of marriage and social support, rather than the specific impact
of reasons for not being married. Those participants who were widowed, single or
divorced scored significantly higher on the CAGE than those who were married (t = 3.79,
P <.001). These individuals were also more likely to report past drug use, X2 (1, N = 681)
= 15.47, j> < .001.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
43
Interest in treatment. Of the 753 individuals who were asked whether they would
like substance abuse treatment. 34% stated that they were interested (n = 253), while 66%
stated that they were not interested (n = 500). Comparisons between those individuals
who expressed interest in treatment and those who did not were made on demographic
and substance abuse measures. Results found that those who expressed interest in
treatment were substantially younger, with a mean age of 65.40 (5.41), as compared to
the age of those who did not want treatment, who had a mean age of 70.27 (7.18), t =
9.50, p < .001. Those who wanted treatment were slightly more educated, with 12.08
(3.14) years of education, compared with 11.58 (2.83) for those who did not want
treatment, t = 2.15, g = .032. Chi-square found that African Americans were no more
likely to express interest in treatment than Whites.
Married individuals were less likely to express interest in treatment than those
who were not married, X2 (1, N = 735) = 40.92, p < .001. This result supported previous
findings by Weisner (1993) and Brennan and Moos (1991) that being married is
negatively associated with treatment seeking.
It was hypothesized that substance abuse severity would be associated with
greater likelihood of treatment interest. This hypothesis was supported. Individuals who
expressed interest in treatment scored significantly higher on the CAGE at 3.01(.95) than
those who did not express interest in treatment, 1.92 (.97), t = 14.60, p < .001.
Chi-square analysis found a significant difference between the two groups on
frequency of drinking, X2 (5, N = 581) = 67.71. Examination of cell sizes in this analysis
revealed a large difference in self-reported daily drinking between those interested in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
44
treatment and those not interested in treatment. Chi-square analysis found this difference
to be significant, X2 (1, N = 581) = 56.76, p < .001. This finding is consistent with
previous findings that daily drinking has been associated with treatment seeking in the
elderly and in younger samples (Brennan & Moos, 1991; Weisner, 1993). There was no
difference between the two groups in current abstention from alcohol.
Chi-square analysis found that those who admitted that they had used illegal drugs
were more likely to express interest in treatment, X2 (1, N = 694) = 71.47, p < .001. See
Table 3, which summarizes differences between those interested in treatment and those
not interested in treatment.
Table 3
Demographics and Severity in Relation to Treatment Interest at Stage 1 (N= 7531
Interested (n = 253) Not Interested (fir = 5001
Mean Age*** 65.40 (5.41) 70.27 (7.18)
Years Education* 12.08 (3.14) 11.58 (2.83)
Percent Married*** 17.7% 39.9%
Percent
White 44.4% 51.3%
CAGE Score*** 3.01 (0.95) 1.92 (0.97)
Percent Drug Use*** 45.7% 19.0%
Percent
Daily Drinking*** 45.6% 16.8%
Percent
Abstainers 35.0% 42.9%
*g< .05. ***p< .001.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
45
Logistic regression analysis at stage 1 examined the utility of demographic factors
and measures of alcohol problem severity in predicting interest in treatment. Age, marital
status, and education were entered at the first step of the model, and CAGE score and
past drug use were entered at the second step. At the first step, age and marital status were
significantly associated with expressing interest in treatment, while education was not.
Together, the demographic factors correctly classified 69.85% of the cases, R2 = .16, X2
(4, N = 660)= 114.87, P < .001. At the second step, the effects of age and marital status
were reduced, though still significant, and relatively stronger effects were found for
CAGE score and past drug use. This second step correctly classified 78.42% of the cases,
with the model predicting significantly more variance than the first step, Step X2 (2, N =
660) = 134.42, p < .001. The model as at the second step was significant, R2 = .32, X2 (6,
N = 660) = 249.26, p < .001. Residual values were also calculated, in order to examine
the amount of prediction error in the equation. For this analysis, standardized residual
values ranged from -4.46 to 3.25, with a mean of 1.90 (.96), suggesting a small amount of
prediction error in the equation (Allison, 1999). The distribution of residuals indicates no
problematic outliers. See Table 4 for results of the regression analysis at stage 1. Figure 8
shows the factors in the model predicting interest in treatment, including regression
weights.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 4
Summary of Hierarchical Regression Analysis for Variables Predicting Expressed
Interest in Substance Abuse Treatment at Stage 1 fN - 6601
Variable B SEB
G
Step 1
Age -0.11 0.02 - 25***
Education 0.03 0.03 .00
Married -1.16 0.21
1 g***
Step 2
Age -0.08 0.18 -.16***
Education 0.05 0.04 .01
Married -1.00 0.24 15***
CAGE Score 0.93 0.10
24***
Drug Use 0.99 0.21
.16***
***2 < -001 .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Education
47
Figure 8. Path analysis for stage 1 results, showing standardized regression
coefficients (N = 660).
C N P
0 0
^_____ ; _______________s
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
D ru g Use
Figure 8 also illustrates the mediational hypotheses regarding stage 1 results. One
of the mediational hypotheses of the study was that the relationship of age to treatment
seeking would be mediated by alcohol problem severity as indicated by the CAGE. A
partial mediational effect was supported (younger age found to be associated with
treatment interest rather than older age). Mediation was examined using the method of
Baron and Kenny (1986), and is subject to the methodological concerns described by
Earleywine (M. Earleywine, personal communication, August 21, 2000). The mediational
effect of CAGE was supported, because the coefficient of treatment interest regressed on
age (.27), was greater than the effect of age when treatment interest was regressed on both
CAGE and age (.21).
Significance of the indirect effect of age was tested, calculating the effects of the
regression coefficients of age predicting CAGE (.25) and CAGE score predicting
treatment interest (.31) in relation to standard error, using the method of Sobel (1982).
Results found that the indirect effect was significant, (3 = .10, g = .025.
It was also hypothesized that the effect of age on treatment interest would be
mediated by the effect of drug use. The partial mediational effect of drug use was
supported because the coefficient of treatment interest regressed on age (.27) was greater
than the effect of age when treatment interest was regressed on both drug use and age
(.23). Significance of the indirect effect of age was tested, calculating the effects of the
regression coefficients of age predicting drug use (.21) and drug use predicting treatment
interest (.15) in relation to standard error. Results found that the indirect effect was
significant, (3 = .03, g = .037.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
49
It was hypothesized that the effect of age on treatment interest would also be
mediated by the effect of education. However, treatment interest regressed on education
was not significant.
Because interest in treatment was associated with CAGE score as well as with
marital status, it was hypothesized that the effect of marital status on treatment interest
would be mediated by CAGE score. A partial mediational effect of CAGE score was
supported, because the coefficient of treatment interest regressed on marital status (.20)
was greater than the effect of marital status when treatment interest was regressed on both
CAGE and marital status (.17). Marital status was also tested as a possible moderator, but
results were not significant.
Significance of indirect effect was tested, calculating the effects of the regression
coefficients of marital status predicting CAGE (.11) and CAGE score predicting
treatment interest (.38) in relation to standard error. Results found that there was a
significant indirect effect of marital status on treatment interest, mediated by CAGE
score, j3 = .05 p = .028.
It was also hypothesized that the effect of marital status would be mediated by the
effect of drug use. This hypothesis was supported, because the coefficient of treatment
interest regressed on marital status (.21) was greater than the effect of marital status when
treatment interest was regressed on both variables (.19). Significance of the indirect effect
was tested, calculating the effects of the regression coefficients of marital status in
predicting drug use (.09) and drug use predicting treatment interest (.15) in relation to
standard error. Results found that the indirect effect was not significant.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
50
Stage 1 path model. Figure 8 illustrates the results of the path model at stage 1,
with significance level of the regression coefficients indicated. Because age was
negatively associated with CAGE, drug use, education and being single, and all four of
these factors were positively associated with treatment interest, it was hypothesized that
these factors would explain the association of age with treatment seeking. However, there
was a significant negative association of age with treatment interest, even after these four
factors were controlled, r = .-15, p < .001. This indicates that these factors did not entirely
explain the effect of being younger on treatment interest.
Stage 2 Results
The second stage of the analysis examined those factors that differentiated
individuals who participated in further pre-treatment evaluation from those who did not.
Comparisons were made on the same demographic and substance use measures employed
at the first stage analysis. See Table 5 regarding demographic composition of the group
that participated in evaluation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
51
Table 5
Demographic Composition of Individuals who Participated in Evaluation (N = 2501
Mean Age
Mean Years Education
65.96 (6.03)
11.92 (3.05)
Ethnic Group n Percent
African American 127 50.8
White 101 40.4
Hispanic 18 7.2
Other 2 0.8
Missing 2 0.8
Marital Status
a
Percent
Married 55 22.0
Remarried 1 0.4
Widowed 34 13.9
Separated 16 6.4
Divorced 113 45.2
Never Married 26 10.4
Missing data 5 2.0
Of the 855 individuals who scored 1 or higher on the CAGE at stage 1, a total of
250 individuals participated in further evaluation prior to treatment entry. However, this
group was not composed of exactly the same individuals who had said they were
interested in treatment at the time of screening (n = 253): some individuals who said they
wanted treatment did not show up, while others who had said they did not want treatment
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
52
showed up in spite of having initially refused. 60.8% of those who showed up to further
evaluation had indicated at screening that they were interested in treatment (n = 152).
23.2% had said that they were not interested in treatment (n = 58). However, through
further contacts with hospital staff, including outreach on the part of treatment staff and
re-admissions to the hospital, these individuals later expressed interest in treatment and
agreed to be evaluated. (Unfortunately, records regarding such contacts were not kept, so
their impact on recruiting for treatment cannot be measured. However, treatment staff
made an effort to stay in contact with all individuals they believed needed alcohol
treatment services, in an unsystematic fashion). There were also 40 individuals who had
missing screening data regarding whether or not they wanted treatment, who showed up
for evaluation (16.0%). See Table 6. See also Figure 3 (above), which illustrates which
individuals entered and exited the process of evaluation and treatment entry.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 6
Age and CAGE scores of those who Participated in Evaluation in Relation to Expression
of Interest in Treatment
Participated in Evaluation
Yes No Total
Yes Age
CAGE
65.0
3.1
(n = 152)
65.9
2.93
(n = 101)
65.4
3.0
(n = 253)
Wanted
Treatment
No Age
CAGE
67.4
2.2
(n = 58)
70.6
1.9
(n = 442)
70.3
1.9
(n = 500)
Missing Age
CAGE
67.5
2.6
(n = 40)
68.3
1.9
(n = 62)
68.3
2.2
(n = 102)
Total Age
CAGE
66.0
2.8
(n - 250)
69.7
2.06
(n = 605)
68.6
2.27
(N = 855)
Differences between these components of the stage 2 sample were explored.
Within this group of 250 people, those who had previously refused treatment and those
who had previously expressed interest at screening were different on both demographic
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
54
and severity measures. Those who had initially refused treatment were older, t = 3.63, g =
.008, had less education, t = 5.65, g - .001, and scored lower on the CAGE, t = 5.12, g <
.001. Drug users were less likely to initially refuse treatment, X2 (1, N = 186) = 7.59, p =
.006. A chi-square analysis found no difference between the two groups by race.
However, there was a significant difference by marital status, with married individuals
more likely to initially refuse treatment, X2 (1, N = 207) = 5.44, g = .02. Although it is
not known what other factors may have influenced individuals to enter treatment who had
initially refused, the results above might be interpreted as a measure of enthusiasm for
treatment-those who said they wanted treatment and did in fact show up might have been
more eager for treatment; those who said they did not want treatment initially, yet still
showed up might have felt reluctance. These results showed the same pattern as the stage
1 analyses, suggesting that older age, less education, being married, and lower self-
reported substance abuse severity were associated with greater reluctance in seeking
treatment.
Dropouts. Just as some individuals who said they did not want treatment showed
up anyway, there were 101 individuals who said they wanted treatment, but did not
participate in further evaluation. Comparing these individuals to those 152 who did
continue, there were no differences by age, education, race, marital status, CAGE score,
or drug use.
Evaluation versus no evaluation. The stage 2 analysis included all 855 people who
scored 1 or higher on the CAGE at the time of screening. It compared those individuals
who showed up for evaluation, versus those who did not show up (regardless of what they
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
had initially said about whether or not they wanted treatment). In other words, while the
stage 1 analysis examined individual differences based on what people said about
wanting treatment, stage 2 examined individual differences based on what they did to
seek treatment.
Those individuals who participated in further evaluation (n = 250) had a mean age
of 65.96 (6.03), which was significantly younger than the age of those who did not
participate (n = 605), who had a mean age of 69.68 (7.25), t = 7. 34, £ < .001. There was
no significant difference in education. Chi-square analysis found that African Americans
were more likely than Whites to participate in evaluation, X2 (1, N = 767) = 12.91, £ =
.001. Married individuals were less likely to participate in evaluation, X2 (1, N = 821) =
14.65, £<.001.
Severity. Those who were evaluated scored significantly higher on the CAGE,
with a mean score of 2.80 (1.09) for those evaluated, compared to 2.06 (1.03) for those
not evaluated, t = 9.50, £ < .001. As at stage 1, chi-square found that those who reported
daily drinking were more likely to be evaluated, X2 (1, N = 591) = 8.55, £ = .003. Those
who reported abstention from alcohol were no more or less likely than those who did not
report abstention. Those with a past history of illegal drug use were more likely to be
evaluated than those who admitted no past illegal drug use, X2 (1, N = 708) = 19.61, £ <
.001. See Table 7 regarding demographic and substance abuse measure differences
between those who were evaluated and those who were not evaluated.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
56
Table 7
Demographics and Severity in Relation to Evaluation at Stage 2 (N= 8551
Evaluated O n = 250) Not Evaluated fn = 6051
Mean Age*** 65.99 (6.03) 69.69 (7.03)
Years Education 11.92 (3.05) 11.73 (2.89)
Percent Married*** 23.3% 37.0%
Percent White** 40.6% 52.5%
CAGE Score*** 2.80(1.09) 2.06(1.03)
Percent
Daily Drinking** 36.7% 24.9%
Percent
Abstainers 44.6% 37.2%
Percent Drug Use*** 43.2% 26.0%
* * * £ < . 001.
Results of regression analysis. A regression analysis using the same predictor
variables as those used at stage 1 was performed. Race was also included, since chi-
square analysis had found a significant difference between African Americans and Whites
in treatment evaluation participation. The dependent variable was the individual’s
evaluation attendance (“had evaluation,” a variable created to indicate evaluation
attendance), as opposed to the expression of interest in treatment used as the dependent
variable in stage 1. Age, marital status, race and education were entered at the first step of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
57
the model, and CAGE score and past drug use were entered at the second step. At the first
step, age, race and marital status were significant. The model correctly classified 71.79%
of cases, R2 = .10. This model was significant, X2 (4, N = 670) = 73.87, p < .001. At the
second step, age remained significant, while marital status and race were no longer
significant. CAGE score was significant, but drug use was not significant. These results
indicated that the effects of race and marital status were not significant, independent of
the effects of CAGE. At the second step the model correctly classified 74.63% of cases,
with the model predicting significantly more variance than the first step, Step X2 (2, N =
670) = 51.93, p < .001. The model at the second step was significant, R2 = .17, X2 (6, N =
670) = 125.06, P < .001. For this analysis, standardized residual values ranged from -4.02
to 1.84, with a mean of 0.09 (.97), indicating a small amount of prediction error in the
equation. The distribution of residuals indicates no problematic outliers. See Table 8,
which shows results of the regression analysis. Also see Figure 9, which shows the
relationship of the factors in the model.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
58
Table 8
Summary of Hierarchical Regression Analysis for Variables Predicting
Participation in Evaluation IN = 670)
Variable B SEB
£
Step 1
Age -0.10 0.02 - 23***
Education 0.01 0.03 .00
Married -0.56 0.21
_ ,09**
Race 0.08 0.04 .05*
Step 2
Age -0.09 0.17
_ ^g***
Education 0.01 0.03 .00
Married -0.40 0.22 -.04
Race 0.05 0.04 .00
CAGE Score 0.62 0.09
25***
Drug Use 0.16 0.21 .00
**gi < .01. ***p < .001
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
59
Figure 9. Path analysis for stage 2 results, showing standardized regression coefficients
(N = 670).
O O
o
00
o
ON
v ' ^
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
D ru g Use
60
The same mediational hypotheses tested at stage 1 were also tested at stage 2. It
was hypothesized that CAGE would mediate the effect of age on evaluation attendance.
A partial mediational effect of CAGE was supported, because the coefficient of
evaluation regressed on age (.23 ), was greater than the effect of age when evaluation
attendance was regressed on both CAGE and age (.18). A significant indirect effect of
age was found, (3 = .06, p = .016. As at stage 1, regression analysis results indicated that
the negative relationship of age to treatment seeking was not explained by the negative
relationship of age to severity.
It was also hypothesized that the effect of age would be mediated by drug use and
education. However, the effects of drug use and education were not significant, so these
hypotheses were not supported.
It was also hypothesized that CAGE score would mediate the effect of marital
status on evaluation attendance. A small partial mediational effect of CAGE score was
supported, because the coefficient of evaluation attendance regressed on marital status
(.11) was greater than the effect of marital status when evaluation attendance was
regressed on both CAGE and marital status (.08). A test for moderator effects of marital
status yielded insignificant results. A significant indirect effect of marital status was also
found, { 3 = .04, p = .023.
Stage 2 path model. Figure 9 illustrates the results of the path model at stage 2,
with significance level of the regression coefficients indicated. As at stage 1, the stage 2
path analysis sought to explain the effect of age on the dependent variable. Because age
was negatively associated with CAGE, drug use, education and being single, and all four
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of these factors were positively associated with evaluation participation, it was
hypothesized that these factors would explain the association of age with participation.
However, there was a significant negative association of age with treatment interest, even
after these four factors were controlled, r = .-18, g < .001. In contrast to the stage 1
model, the stage 2 model found that drug use did not significantly predict treatment
seeking. The stage 2 model also illustrates that the effect of being married was not a
significant predictor once the other variables were controlled.
Stage 3 Results
The third stage of the analyses examined the measures given at the time of
evaluation, to differentiate individuals who began treatment from those who did not. The
second proposed model (the developmental distress model) was also tested (Figure 2). As
noted above, the measures used at this stage of evaluation were not administered
consistently to all 250 individuals, due to inconsistent evaluation procedures and staff
fluctuations, including layoffs at the Veteran’s Administration Hospital. For this reason,
the results based on these measures must be considered preliminary. Measures
administered prior to treatment entry included the MAST- G (n - 160), The Folstein
MMSE (n = 90) HRS-A (n = 84) and the HRS-D (n = 81). There were 45 individuals who
completed all four measures, while 205 completed some of the measures but not all. See
Table 9, which shows the correlation coefficients between the variables examined at
stage 3.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
62
Table 9
Correlation of Scale Scores
Age Educ. CAGE MAST-G HRS-A HRS-D MMSE
Age
Correlation 1.0
Significance
n 855
Education
Correlation
Significance
n
CAGE
Correlation
Significance
n
MAST-G
Correlation
Significance
n
HRS-A
Correlation -.01 .087 .17 .19 1.00
Significance
n 84 80 84 51 84
HRS-D
-.16 1.00
.000
794 794
-.23 .01 1.00
.000 .724
855 794 855
-.14 .15 .43 1.00
.084 .066 .000 .00
160 152 160 160
-.01 .087 .17 .19
.920 .443 .126 .182
84 80 84 51
Correlation -.14 .08 .20 .38 .81 1.00
Significance .309 .507 .071 .007 .000
n 81 77 81 49 78 81
MMSE
Correlation -.34 .33 .14 .37 .07 .12 1.00
Significance .001 .002 .186 .007 .53 .304 .
n 90 86 90 57 79 76 90
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
63
Differences between those who completed all the measures and those who
completed only some of them were explored. Results found no significant differences in
age or CAGE score between those who completed the all evaluation measures and those
who did not. Chi-square found no significance in drug use, marital status or race between
the two groups.
The MAST-G was completed by 160 individuals. Scores on the MAST-G ranged
from 0 to 24, with a mean score of 10.86 (6.77). A score of 5 or higher on the MAST-G is
considered indicative of a likely alcohol problem. MAST-G score was significantly
correlated with CAGE score, r = .43, g < .001. There was no correlation of MAST-G
score with duration of drinking. One-way ANOVA found no association of MAST-G
score with self-reported frequency of drinking. Contrast analysis also found no significant
association between these variables. There was no difference in MAST-G score between
those who reported using drugs other than alcohol and those who did not.
148 individuals completed a questionnaire regarding duration of alcohol problems
and past treatment history. On this questionnaire, individuals were asked “About how old
were you when you first could have considered yourself to have an alcohol problem?”
Response to this question was subtracted from the participant’s age at the time of the
evaluation to obtain an estimate of the duration of drinking problem. Duration ranged in
length from .25 to 65.75 years, with a mean length of 24.95 years, and a standard
deviation of 15.35 years. The distribution was flattened, with scores spread out across the
entire range. Duration was significantly associated with CAGE score, r = .28, g = .002.
There was no significant correlation of duration with MAST-G score. Positive correlation
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
64
of duration of alcohol problems with age was not quite significant, r = .18, p = .053.
There was no significant association of duration of alcohol problem with race, education,
or marital status.
Of the individuals who completed the questionnaire, 60% had received alcohol
treatment in the past, while 40% had not. It was expected that past treatment would be
associated with measures of alcohol problem severity. Evidence for this association was
mostly positive. There was a significant difference in CAGE score between those who
reported past treatment, with a mean score of 3.21 (.90) as compared to those who
reported no past treatment, who had a mean score of 2.80 (1.02), t = 2.61, p = .010. Those
who reported past alcohol treatment scored a mean of 13.84 (5.70) on the MAST-G,
compared to 10.21 (6.15) for those with no past treatment, t = 2.92, p = .004. However,
there was no significant difference in duration of alcohol problems between those who
had past treatment and those who had not.
It was also hypothesized that past AA attendance would be associated with
severity. Results were mixed. 72% had attended AA meetings in the past, while 28% had
not. A significant difference on the MAST-G was found for past AA attendance, with a
mean score of 13.02 (5.82) for those who attended, versus 8.58 (5.55) for those who had
not, t = 4.08, p < .001. However, there was no difference in CAGE score between those
who had attended AA and those that had not, nor was there a difference in duration
between those who had attended AA and those who had not.
Relationship of demographics to severity measures. At stage 1, older age was
associated with lower levels of alcohol problem severity. At stage 3, this association was
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
65
not apparent, possibly due to smaller number of subjects and fewer very old subjects.
While the range of ages at stage 3 was 55 to 89, which was similar to that at stage 1 (55 to
91), there were fewer individuals at the older end of the age range at stage 3. For
example, 18 percent of the sample were over 75 at stage 1, while only 8 percent were
over 75 at stage 3. As noted above, individuals who participated in further evaluation
were significantly younger than those who did not. At stage 3 there was no correlation of
age with CAGE or MAST-G. There was no significant association of age with past AA
attendance or alcohol treatment. There was no association of age with past drug use.
Analysis of the effects of race focused on differences between African Americans
and Whites, since these two groups combined made up 91% of the sample. There was no
significant difference between the two groups in age. Whites had more education with
13.11 (2.42) years, versus 11.15 (3.17) for African Americans, t = 5.07, p < .001. Chi-
square found no significant difference in marital status between African Americans and
Whites.
In contrast to stage 1, which appeared to show a clear relationship of race and
severity, the association at stage 3 was not uniform. Whites scored significantly higher on
the MAST-G, with a mean score of 12.46 (6.07), versus 9.81 (6.79) for African
Americans, t = 2.43, p = .02. This was the reverse of the stage 1 results, which found
higher CAGE scores for African Americans as compared to Whites. There was no
significant difference between the two groups on the CAGE, HAM-A, HAM-D, or the
Folstein MMSE. Chi-square found no significant difference between African Americans
and Whites in self-reported illegal drug use.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
66
Marital status. There was no significant difference on the Folstein MMSE, HAM-
A, HAM-D between those who were married and those who were not married. Married
individuals scored slightly lower on the MAST-G, with a mean score of 8.78 (7.52),
compared to those who were either never married, divorced, separated or widowed, who
scored 11.35 (6.34), t = 2.06, p = .041. There was no significant difference between the
two groups on the CAGE.
Depression. The HRS-D was used to measure depression level. 81 individuals
were given the HRS-D. Scores ranged from 0 to 35, with a mean score of 9.61 (7.74).
While this scale is generally used as a measure of severity rather than as a diagnostic
instrument, a score of 10 is considered indicative of at least mild depression among older
adults (Scogin, 1994). Depression was not significantly associated with demographic
measures. There was no association between age, marital status or race with depression.
It was expected that depression would be significantly correlated with alcohol
problem severity in the sample. Results were mixed. Depression was significantly
associated with alcohol problem severity as measured by the MAST-G, r = .38, p = .007.
Correlation with CAGE score was not significant. However, the correlation was in the
same direction; significance might have been found with a larger sample size r = .20, p =
.071. Neither past alcohol treatment nor AA attendance was significantly associated with
depression. Depression was also not significantly associated with duration of alcohol
problems.
Anxiety. The Hamilton Anxiety Scale (HRS-A) was completed by 84 individuals.
Scores ranged from 0 to 25, with a mean score of 7.11 (5.71). Cutoff level for significant
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
67
anxiety is 14. Results suggest that the sample in the current study was not suffering from
clinically significant anxiety. Anxiety level was not significantly associated with age,
education, race or marital status.
HRS-A was not significantly associated with either the CAGE or the MAST-G. It
was also not correlated with duration of alcohol problems, past treatment or participation
in AA. However, the HRS-D and HRS-A scores were highly correlated, r = .81, g < .001.
Folstein MMSE. The Folstein MMSE was used as a dementia screening device,
and was completed by 90 individuals. Two individuals, who scored 11 and 16, were
eliminated from the sample (including stage 1 and 2 analyses), because these scores
indicated severe cognitive impairment that would most likely have precluded them from
taking part in group treatment. With those deletions, scores on this test ranged from 18 to
30, with a mean score of 25.39 (3.15). This score is at the 14th percentile, for adults aged
65 to 79 with more than 9 years of education (Tombaugh, McDowell, Krisjansson &
Hubley, 1996). These results suggest that the average individual being evaluated suffered
from a mild degree of cognitive impairment. As expected, higher age was significantly
associated with a lower score on the MMSE, r = -.34, g = .001. Higher education level
was associated with higher scores on the MMSE, r = .33, g = .002, a finding consistent
with previous studies, (Magaziner, Bassett, & Hebei, 1987). See Table 10 for more
detailed description of MMSE scores of individuals in the sample, with percentile ranges
and corresponding impairment level.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
68
Table 10
Scores on the Folstein Mini-Mental Status Exam, in Relation to Published Norms In =
90}
Score Range n
Norms in
Percentiles Cognitive Status
18-20 5 1-4% moderately impaired
21-24 30 4-9% mildly impaired
25-27 26 12-29% low average
28-30 29 41-86% average to high average
Source for percentile norms: Tombaugh et al., 1996.
It was expected that lower scores on the MMSE would be associated with higher
scores on measures of alcohol problem severity. However, higher scores on the MMSE
were significantly associated with higher scores on the MAST-G, a surprising finding, r =
.37, p = .005. There was no significant correlation between the CAGE or duration of
alcohol problems and the MMSE. There was no significant relationship of MMSE scores
to measures of depression or anxiety.
Enrollment in treatment. Of the 250 individuals included in the stage 3 analysis,
124 began treatment, while 126 did not. In contrast to the stage 1 and stage 2 analyses,
there was no significant difference in age between the two groups. Those who entered
treatment were slightly younger, but this difference was not significant. Chi-square found
no significant difference between African Americans and Whites in treatment enrollment.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
69
There was also no difference by marital status. Education was significantly higher for
those who enrolled in treatment, with a mean of 12.41 (3.01) years, compared to 11.39
(3.04) years for those who did not enroll, t = 2.61, p = .010. See Table 11. Those in
treatment also scored higher in education when compared to the individuals from the
original sample who never entered treatment (n = 649), who had a mean score of 11.67
(2,87), t = 2.63, p = .009.
Table 11
Demographics in Relation to Treatment Entry (N- 250)
Enrolled In = 124) Not Enrolled fn = 126)
Mean Age 65.79 (5.76) 66.13 (6.29)
Years Education** 12.41 (3.01) 10.39 (3.04)
Percent Married 23.4% 21.4%
* * P < .001.
It was hypothesized that those who enrolled in treatment would score significantly
higher on measures of alcohol problem severity. Results confirmed this hypothesis. Those
who enrolled in treatment scored higher on the CAGE, with a mean score of 3.06 (.93),
compared to those who did not enroll, 2.54 (1.18), t = 3.91, p < .001. Those in treatment
also scored significantly higher than those individuals from the original sample who
never entered treatment (n = 731) who had a mean score of 2.12 (1.06), t = 9.82, p < .001.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Those in treatment also scored significantly higher on the MAST-G, with a mean score of
12.23 (6.25) for those who enrolled, compared to 9.01 (7.02) for those who did not enroll,
t = 2.99, g = .003. Chi-square found no association between drug use and treatment
enrollment. See Table 12 regarding results of substance abuse measures of those who
enrolled in treatment versus those who did not enroll.
Table 12
Enrolled fn = 1241 Not Enrolled fn = 1261
CAGE Score*** 3.06 (.93) 2.54(1.18)
(n = 250)
MAST-G Score**
(n = 160)
12.23 (6.25) 9.01 (7.02)
Percent Drug Use 38.7% 27.8%
**g< .01. ***g< .001.
It was hypothesized that psychological distress would motivate those with an
alcohol problem to seek treatment. Results were mixed. Scores of those who enrolled
were higher on the HRS-D, with a mean of 11.79 (7.52) for those who enrolled,
compared to 7.54 (7.42) for those who did not enroll, t = 2.56, g = .012. These mean
scores indicated that the average person who enrolled in treatment met the cutoff for
clinically significant depression, while the average person who did not enroll did not
meet this cutoff (a score of 10, Scogin, 1994). There was no significant difference in
anxiety between those who enrolled in treatment and those who did not enroll.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
71
Adequate cognitive functioning was proposed as a facilitating factor in treatment
entry for older adults, so it was hypothesized that those who entered treatment would
score higher on the Folstein MMSE than those who did not. Those who enrolled in
treatment did in fact score higher on the Folstein, with a mean score of 26.97 (2.87),
compared to those who did not, who scored 24.11 (2.85), t = 4.45, g <.001. The mean
score of those who enrolled lies at the 29th percentile, in the low average range, while the
score of those who did not enroll lies at the 9lh percentile, in the mildly impaired range
(Tombaugh et al., 1996). These results indicate that those who enrolled in treatment were
significantly less impaired than those who did not enroll. See Table 13.
Table 13
Anxiety. Depression, and Cognitive Function in Relation to Treatment
Enrollment at Stage 3
Enrolled fn = 1241 Not Enrolled fn = 1261
8.00 (4.83) 6.14(6.45)
11.57 (8.21) 6.91 (6.64)
26.97 (2.87) 24.11 (2.85)
*g < .05. ***g < .001.
Correlations between the variables at stage 3 are illustrated in Table 9. Because of
the large amount of missing data from the pre-treatment evaluation measures, small
sample size hampered testing a logistic regression model that incorporated all the factors
HRS-A Score
(n = 84)
HRS-D Score*
(n = 81)
Folstein Score***
(n = 90)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
72
of the second model. In order to increase the sample size at stage 3 for the purpose of
logistic regression analysis, the use of imputation methods to estimate missing data was
explored. In particular, it was desirable to estimate, if possible the HRS-D and MMSE
scores of those individuals who did not complete these measures. Imputation based on
correlation of these measures with non-missing variables was not an acceptable solution,
because correlations were too small and the amount of missing data too large (Cohen &
Cohen, 1975; Timm, 1970). Therefore, those individuals who did not complete both the
HRS-D and the MMSE were dropped from the regression analysis.
The model tested at stage 3 examined the relative importance of age, marital
status, education, CAGE score, cognitive status and depression in predicting treatment
entry. CAGE score was used as a measure of alcohol problem severity, since it was
completed by more participants than the MAST-G. In a hierarchical model, age and
marital status were entered at the first step, education and MMSE score were entered
second, and CAGE score and HRS-D were entered third (n = 70). First step results found
that age was significant, while marital status was not. The model correctly classified
61.43% of the cases, R2 = .10, X2 (2, N = 70) = 7.41, p = .025. At the second step, age
was no longer significant, while MMSE score was significant. The model correctly
classified 71.43% of cases, R2 = .23, X2 (4, N = 70) = 18.31, p = .001. Step chi-square
indicated that the second model was significantly better than the first, step X2 (2, N = 70)
= 10.90, p = .004. At the third step, only MMSE and CAGE score were significant. See
Table 14.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
73
Table 14
Summary of Hierarchical Regression Analysis for Variables Predicting
Treatment Entry IN - 70)
Variable B SEB
Step 1
Age -0.10 0.05 -.21*
Married -0.59 0.59 .00
Step 2
Age -0.08 0.05 -.08
Married -0.26 0.64 .00
Folstein MMSE 0.26 0.10 , 22*
Education 0.13 0.09 .00
Step 3
Age -0.04 0.06 .00
Married -0.50 0.75 .00
Folstein MMSE 0.28 0.11
24**
Education 0.19 0.11 .12
CAGE 0.92 0.33 .27**
HRS-D 0.06 0.04 .03
*£<.05. **£<.01.
The model correctly classified 78.57% of cases, R2 = .37, X2 (6, N = 70) = 32.14,
£ < .001. Step chi-square indicated that the third model was significantly better than the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
second, step X2 (2, N = 70) = 13.83, p = .001. For this analysis, standardized residual
values ranged from -1.54 to 4.41, with a mean of 0.07 (1.04), which indicates a small
amount of prediction error in the equation. The distribution of residuals indicates no
problematic outliers. Results of this logistic regression analysis indicated that the need
factor of CAGE score, and the facilitating factor of cognitive functioning were the most
important factors in predicting treatment entry, and that these two variables explained the
relationship of age to treatment entry. The significantly higher R2 at the second and third
stages, which incorporated elements of the proposed developmental distress model,
suggested that this new model was a promising explanation for treatment entry.
Specific mediational hypotheses were also tested. Because the effect of emotional
distress on treatment entry had been hypothesized to explain the effect of alcohol problem
severity on treatment entry, a hierarchical regression model was tested, using just these
variables. This model had CAGE at the first step and HRS-D score at the second step (n=
81). Results found that when it was entered alone, CAGE was a significant predictor at
the first step, with 77.78% of cases correctly classified, X2 (1, N = 81) = 22.33, p < .001.
The addition of HRS-D at the second step was not significant. See Table 15. While these
results must be very tentative given the large percentage of missing data, findings did not
support the hypothesis that depression could explain the effect of alcohol problem
severity on treatment entry.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
75
Table 15
Summary of Hierarchical Regression Analysis for CAGE and HRS-D in Predicting
Enrollment in Treatment tN = 811
Variable B SEB
Step 1
CAGE Score 0.91 0.24
24***
Step 2
CAGE Score 0.88 0.24
34***
HRS-D 0.07 0.04 .13
* * * £ _ < .01.
The possibility that depression might function as a moderator of the effect of
CAGE score on treatment entry was also considered. According to Barron and Kenny
(1986), when an independent variable (CAGE score ) and the proposed moderator (HAM-
D) are continuous variables, the effect of moderation is tested by adding the product of
the moderator and the independent variable to the regression equation, and then running
the regression analysis to see if the product is significant when the effects of the
moderator and the independent variable are controlled. This procedure was conducted,
with non-significant results, indicating that depression did not function as a moderator in
the relationship of alcohol problem severity to treatment entry.
It was also hypothesized that the relationship of alcohol problem severity to
treatment entry would be mediated by the effect of cognitive impairment. There was a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
significant association between MMSE score and treatment entry. However, as noted
above, there was no significant correlation of the MMSE with the CAGE, while the
correlation between the MMSE and the MAST-G was in the opposite direction from that
which was expected. Therefore, no support for this hypothesis was found.
Stage 3 path analysis. The stage 3 path analysis results are illustrated in Figure 10
(n = 70). Significant pathways were found between age and CAGE score, and CAGE
score and treatment entry, and MMSE score and treatment entry. Education was also
significantly associated with MMSE score. Some of the other paths, such as the one
between age and MMSE score, would most likely have been significant with a larger
sample size, since the correlation coefficient for these variables were significant. This
model highlights the importance of MMSE score as a facilitating factor in predicting
treatment entry for older adults.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Education
77
Figure 10. Path analysis for stage 3 results, showing standardized regression coefficients
(N = 70).
o
o
o
©
O
V ._______________________ J
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Depression
78
Discussion
Treatment Seeking Behaviors in Three Stages
The current study examined the utility of a proposed developmental distress
model, a modified version of the behavioral health services utilization model, in
explaining alcohol treatment seeking in a sample of older male veterans (Andersen &
Newman, 1973). In doing so, it used a different method than that of previous studies.
Prior studies have examined predictors of treatment entry by comparing those individuals
in alcohol treatment to individuals with an alcohol problem who were not in treatment
(Gomberg, 1995; Weisner, 1993). In these studies, factors examined for their association
with treatment were based on data gathered after participants had already begun
treatment, and were compared to samples gathered from the community. In contrast, the
present study had the advantage of following a single sample, selected using uniform
criteria, through three stages of treatment seeking: expression of interest in treatment,
participation in pre-treatment evaluation, and enrollment in treatment. This longitudinal
study design has been advocated as a promising approach to understanding predictors of
treatment seeking behavior (Hingson et al., 1980). In the current study, measurement of
treatment seeking predictors was conducted prior to the treatment seeking behavior to be
examined. This afforded the present study a distinct advantage over prior studies.
Consistent sampling criteria were used, and longitudinal methodology allowed for
temporal predictors of treatment seeking to be tested. In the discussion below, findings
regarding the significance of the predictors, and their role in the developmental distress
model, is discussed in relation to each of the three stages examined.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
79
Age. Severity and Treatment Seeking
One of the hypotheses examined in this study was a possible association between
greater age and increased severity of alcohol problems, consistent with a developmental
perspective on how alcohol use disorders originate. This developmental view, based on
work primarily with younger and middle aged adults, holds that alcohol problems tend to
get worse with duration (Gomberg, 1980; Jellinek, 1960). Because greater severity has
been consistently associated with treatment seeking, it was proposed that older adults
would be more likely to seek treatment than relatively younger adults whose alcohol
problems were of shorter duration and less severity. This position was supported by the
work of Hingson and associates (1980), who found that adults with an alcohol problem
who were over 50 were more likely to be in treatment than adults with an alcohol
problem who were under 50, in a general population survey. Unfortunately, age range and
distribution was not reported in this study, so the implications of their findings for an
elderly population were uncertain.
The theory that age would be associated with treatment seeking was tentatively
proposed. Because the Hingson study compared adults over 50 to those under 50, it was
thought that his results might have little relevance to age effects within the current
sample, whose ages ranged from 55 to 90, with a mean age of 68. In addition,
epidemiological studies of alcohol abuse and dependance have found that prevalence
rates of alcohol abuse and dependence are lower in older adults than in younger ones.
Results of consumption studies have been consistent with prevalence studies, showing
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
80
lower rates of consumption and higher rates of abstinence among older adults (see
Bucholz, Sheline, & Helzer, 1995, for a review). Based on this body of literature, it was
recognized that age might be negatively correlated with severity in the present sample.
Because greater severity of alcohol problems has been associated with greater
interest in treatment entry, it was proposed that both older age and greater severity would
be associated with interest in alcohol problem treatment (Gomberg, 1995; Weisner,
1993). Because of the strong relationship found in prior studies between severity and
treatment entry, it was also hypothesized that the effect of age would be mediated by
severity. Findings of the current study confirmed only some of these hypotheses: severity
was negatively, rather than positively associated with age; younger age and greater
severity were both associated with treatment seeking; the inverse effect of age was
partially mediated by severity but still had an independent effect on treatment interest and
evaluation, but not in regard to treatment entry per se.
Implications for the Behavioral Model
This study identified factors that may help adapt the behavioral model in
predicting substance abuse treatment seeking among older male veterans, and possibly
other populations of older adults. Factors correlated with actual treatment seeking
behavior at stage 2 and 3 included predisposing factors (younger age, not married), need
(alcohol problem severity, comorbid drug use, depression), and facilitating variables
(cognitive status). As discussed in greater detail below, certain variables remained
significant after controlling for the effects of other variables in the model. In particular,
younger age emerged as the most important predisposing variable, while alcohol problem
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
81
severity was the most important need characteristic. Across all three stages tested in the
study, CAGE score contributed the most to the formulations of the behavioral model that
were tested.
The emergence of need as the most significant contributor to the model at all three
stages was consistent with the original formulation of the behavioral model (Andersen &
Newman, 1973). It was also consistent with findings from other studies that have
examined treatment seeking for medical services among older adults, in which medical
need was more important than demographic factors (Haug, 1981), as well as prior studies
comparing male problem drinkers in treatment to those not in treatment (Weisner, 1993).
Measurement of treatment need in the context of model testing. This study
provided the opportunity to examine treatment need using three alcohol measures-the
CAGE, the MAST-G and frequency of drinking. Correlation between the CAGE and
MAST-G was significant (.43). Both were positively associated with treatment entry.
However, these two alcohol problem scales had differing relationships to the other
variables in the study, raising questions regarding possible construct validity differences
between the two measures. For example, the MAST-G was positively associated with
depression score, though there was no association of CAGE with depression. In contrast,
duration of alcohol problems and past drug use was associated with the CAGE but not the
MAST-G. Because a substantial number of individuals at stage 3 completed the CAGE
but not the MAST-G (36%), interpretations based on these differences should be made
with caution. However, one possible explanation for the difference in association with
depression, drug use and duration of alcohol problems might lie in the greater present-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
82
focus of the MAST-G. This measure was designed as a measure of current alcohol
problem severity, with some modifications to the original MAST scale in order to tailor it
to the older adult population (Blow et al., 1992). The CAGE, in contrast, is a brief
screening measure designed to identify lifetime history of alcohol problems (Blow,
1998). If current severity level could be expected to show a stronger relationship to
current depression than past history of alcohol problems, the disparity in the relationship
of the alcohol scales to depression score is understandable. Likewise, it makes sense that
past drug use and duration of alcohol problems would be associated with the CAGE
rather than the MAST-G, because the MAST-G asks questions primarily about current
behavior and the CAGE addresses alcohol problem history. Past drug use and early
alcohol problems may or may not be associated with a current alcohol problem.
Results suggest that current self-reported frequency of drinking should be
interpreted with caution. Frequency of drinking was related to CAGE score in a
curvilinear fashion. This finding indicated that self-reported abstention from alcohol was
not a good measure of alcohol problem severity: There was no relationship of abstention
to number of alcohol related problems, nor was abstention associated with a lack of
interest in treatment. On the other hand, daily drinking was associated with past alcohol
problems as well as with treatment interest. These results suggest that while daily
drinking may be a useful indicator of treatment need if alcohol-related problems are also
detected, the absence of current drinking does not indicate a lack of treatment need. As
the current results found, those who are not currently drinking, yet had a history of
alcohol problems as indicated by the CAGE, were still interested in treatment. For such
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
83
individuals, it is possible that treatment would be sought in order to maintain sobriety.
Comorbid drug use. Prior studies have not examined the effects of comorbid drug
and alcohol abuse on treatment seeking behavior in the elderly, nor have studies of
younger adults examined this association. While the abuse of prescription medications by
the elderly has increasingly become a focus of concern for clinicians and researchers, very
little is known about the use of illicit drugs such as amphetamines, heroin and cocaine
among older adults (Blow, 1998). Nevertheless, such substance abuse has considerable
prevalence in clinical samples of older male veterans, as the present study demonstrated.
The inconsistent association of drug use with treatment seeking across the 3 stages
examined in the current study leaves the importance of comorbid drug use in treatment
models an open question. It was hypothesized that the presence of drug use other than
alcohol would function as an aspect of substance abuse problem severity, and therefore
that the presence of drug use would increase treatment seeking. The presence of drug use
was associated with the expression of treatment interest, both in chi-square results and in
the stage 1 regression model, though the smaller regression coefficient size at stage 1 for
drug use relative to that of the CAGE indicated that alcohol problem severity was a more
important predictor of treatment interest than drug use. Those who used drugs were more
likely to attend an evaluation at stage 2, in univariate chi-square analysis. However, the
effect of drug use was not significant in the stage 2 regression model with other variables
controlled statistically, and was not significantly associated with actual treatment entry.
This discrepancy suggests that while the presence of drug use may have functioned as an
aspect of substance abuse severity at the time of screening that increased interest in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
84
treatment, the effect of drug use was not as strong at later stages of the treatment entry
process. In the context of the behavioral model, comorbid drug use may be useful in
predicting intentions but not actual behavior of candidates for treatment.
Marital status. Researchers have found that being single has been associated with
elevated levels of distress among older adults, including Whites, African Americans and
Asians (Krause et al., 1995). In the current study, it was expected that being single would
be positively associated with alcohol problem severity, as previous studies of older adults
have found (Bucholz et al., 1995), and that this mediational relationship might explain the
association of marital status with treatment seeking behavior. However, the mediational
effects of CAGE score observed at stages 1 showed that the association of alcohol severity
only partially explained the effect of being unmarried on treatment interest. So even
though being unmarried was associated with greater severity, the effect of severity did not
totally account for the effect of being unmarried on treatment interest. This suggests that
other factors, such as the effect of social support provided in marriage, may be important
in explaining how treatment interest is influenced by marital status.
The first stage regression analysis found that being married was negatively
associated with interest in treatment, independent of the effect of alcohol problem severity
and drug use. These results were consistent with those of prior studies that have found
being single positively associated with participation in treatment for both younger and
older adults (Brennan & Moos, 1991; Weisner, 1993). However, at stages 2 and 3 the
independent effect of marital status was not significant. The lack of significance for the
independent effect of marital status at stage 2 indicated that marital status per se was not a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
85
key factor in the individual’s participation in a pre-treatment evaluation, but was mediated
by CAGE score. At stage 3 there was no correlation between marital status and treatment
entry. The discrepancy between stage 1 results and those of stages 2 and 3 suggest that
marital status predicts interest in alcohol treatment, but not actual treatment seeking
behavior. As with drug use, marital status seemed more related to intentions regarding
treatment than actual behavior. Reasons for this discrepancy are not clear. It is possible
that at the time of screening, being married diminished the perceived need for treatment of
patients in the study, but not to the extent that it affected subsequent behavior.
Ethnicity. Although there was no theoretical reason to expect that there would be
differences in treatment seeking between ethnic groups, the ethnic composition of the
sample in the present study provided an opportunity for exploration. The role of ethnicity
was examined in terms of differences between Whites and African Americans, since
together these groups made up over 90% of the sample. Previous studies of older men
have found that African Americans scored higher on measures of severity as well as on
measures of the social consequences of alcohol dependence (Gomberg, 1995). Consistent
with these findings, the stage 1 sample showed that being African American was
associated with higher CAGE score and greater prevalence of drug use. African Americans
were also less likely to be married. It might have been expected that African Americans
would be more likely to seek treatment, because of these factors. This was the case at stage
2. However, there was no effect for race once CAGE score was entered in the model,
indicating that the effect of race was an artifact of the association of race with severity (see
Table 8). Because the effects of race were not significant once severity was controlled,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
86
race did not seem likely to be an important component of the behavioral model.
The significantly higher scores of Whites than African Americans on the MAST-G
at stage 3 is somewhat difficult to explain. As noted above, the MAST-G lays relatively
greater emphasis on current drinking problems than the CAGE. The current results
indicate that in this sample, African Americans had greater past drinking problems, but
fewer current drinking problems than Whites. It may also be that those African Americans
with a more severe current alcohol problem were less likely to progress to the point of
evaluation than their white counterparts.
Education. The behavioral model includes education level as a predisposing factor
in health service utilization, on the assumption that a higher level of education leads to
greater knowledge of health care systems and greater ease of access. In the current study, it
was hypothesized that education would be positively associated with treatment seeking
behavior, though the effect of education was small in previous studies (Weisner, 1993).
The difference between treatment seekers and non-treatment seekers was small but
significant at stages 1 and 3, and not significant at stage 2. As both of these factors were
negatively associated with age, they provided additional explanation for the negative
association of age with treatment seeking. However, the effects of education on treatment
seeking at all three stages were not significant once the effects of other variables were
controlled, suggesting that positive association of education with treatment seeking at
stages 1 and 2 was an artifact of the negative association of education and age. Therefore,
results of the current study do not support the inclusion of education in the behavioral
model of alcohol problem treatment seeking.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
87
In summary, this investigation of the utility of the behavioral model in predicting
treatment seeking behavior among older male veterans identified similar predictors as
those found in previous studies. Most strikingly, the effect of age was significant
independent of other variables in the models tested at stage 1 and 2 (see Figures 8 and 9).
These findings call into question the extent to which predisposing demographic factors
have been downplayed in previous formulations of the behavioral model, which have
emphasized the central role of need factors in determining treatment seeking (Andersen &
Newman, 1973; Krause, 1990). For older adults, it seems particularly important to explore
the negative effects of age and other factors that may be associated with age, in developing
appropriate models of treatment seeking for substance abuse problems.
The Developmental Distress Model
The current study expanded the behavioral model to encompass factors not
previously tested in an older adult population. The developmental distress model proposed
adaptations to the behavioral model in order to better predict treatment seeking. These
adaptations contained additional variables falling into two categories: need and facilitating
variables. Psychological distress was proposed as a possible motivating factor for all
individuals with an alcohol problem, which could function as a manifestation of treatment
need. Because prior studies have found age-associated increases in depression and anxiety
in association with alcohol problems, distress was especially important to examine.
Cognitive status was proposed as a facilitating factor particularly relevant to older adults
due to the increased prevalence of cognitive impairment with age.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
88
Psychological distress. Both anxiety and depression were measured at the time of
evaluation. It was expected that elevated anxiety would be associated with higher levels of
alcohol problem severity and with greater interest in treatment. However, scores on the
Hamilton Scale for Anxiety (HRS-A) were not significantly associated with any
demographic or severity measures, nor was anxiety associated with treatment entry. This
lack of significance may partly be attributable to small sample size and low statistical
power that resulted. It may also have been due to the fact that the mean anxiety score of
the sample as a whole was quite low, approximately half of the clinically significant HRS-
A score.
In contrast to anxiety, depression was significantly associated with alcohol
problem severity and with treatment entry in the current study. The finding that those who
entered treatment scored higher on a measure of depression than those who did not enter
treatment was consistent with results obtained by Brennan and Moos (1991), Chan and
associates (1991), and Woodruff (1973). While these studies found that adults in treatment
were more depressed than a comparable sample who were not in treatment, the current
study was the first known study to show that depression level predicts future treatment
entry. However, the hypothesis that depression would mediate the effect of severity on
treatment entry was not supported. The effect of depression was not significant in the stage
3 regression model, in which it explained no additional variance over alcohol problem
severity, suggesting that depression associated with treatment entry was linked to higher
alcohol problem severity.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In spite of these negative results, depression should not be wholly discounted. Lack
of significance may have been due in part to small sample size in the stage 3 regression
analysis. In addition, the significant association of elevated depression level with treatment
entry has implications for both screening and planning treatment for substance abuse
among older veterans. Because of the significant association of depression with alcohol
problem severity as measured by the MAST-G, the presence of depression at screening
may indicate a possible alcohol problem. In such screening situations, it is possible that
the comorbid presence of depression and alcohol problems may provide an additional
avenue for intervention. Prior studies, as well as the current one, have shown that rates of
treatment refusal for alcohol problems are high (Stephan, Swindle & Moos, 1992). Stigma
around the presence of an alcohol problem is also high (Blow, 1998). An offer of
treatment for depressed mood may be a more acceptable alternative, and may provide an
inroad to eventually addressing the alcohol problem as well.
The significant association of depression with severity and with treatment entry
raises interesting problems for treatment planning. On the one hand, because of the
association of alcohol problem severity and depression, and the higher rates of depression
for those in treatment, it is clear that alcohol treatment must also strive to improve mood.
This recommendation has been voiced by many clinicians working with younger adults as
well as older adults, and is supported by the results of the present study (Blow, 1998;
Schonfeld & Dupree, 1991). Depression has been identified as a precipitating factor in
episodes of problem drinking among older males in treatment (Schonfeld & Dupree,
1991). Studies of younger adults have identified more severe craving for alcohol
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
90
associated with greater depression among individuals in treatment (Velasquez et al.,
1999). This seems to imply that effective treatment for depression has the potential to
reduce future episodes of problem drinking. On the other hand, if depression helps to
motivate individuals to enter treatment, it does seem possible that improving mood might
lead to treatment dropout. Therefore, it seems necessary that treatment for comorbid
depression in a substance abuse intervention would strive to avoid this loss of motivation,
possibly by emphasizing the importance of ongoing treatment to sustain improvement in
mood. This is an intriguing problem that deserves greater attention in future studies.
Cognitive status. This is the first study to find a positive association of cognitive
functioning with treatment entry in an elderly sample. This effect was independent of the
effects of age, alcohol problem severity and education, a finding that supports the
inclusion of cognitive status as a facilitating factor in the developmental distress model
(Figure 10). It may seem obvious that cognitive impairment would negatively affect the
ability of older adults to take part in an outpatient treatment program. However, this
finding highlights an important concern in substance abuse treatment planning for older
adults-how to make alcohol treatment accessible and effective in a population whose
ability to participate in traditional forms of talk therapy may be compromised by memory
loss and other decrements in cognitive functioning.
In the present study, both those who entered treatment and those who did not
showed some evidence of cognitive impairment. The significant Folstein MMSE score
difference between these two groups suggests that treatment barriers exist for those with
relatively greater impairment. Modifications to screening and outreach procedures may
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
91
need to take this into account in order to facilitate treatment entry. At a minimum, this
would include active follow-up on the part of treatment staff to help impaired individuals
to remember appointments and to locate the clinic where therapy sessions are held.
Involvement of family members, if available, also seems indicated in order to access
treatment more effectively.
Because those in treatment were in the low-average range on the MMSE for their
age group, results suggest that modifications to treatment procedures may also be needed
in order to accommodate older adults with reduced cognitive ability. No studies to date
have examined whether impaired older adults can benefit from group or individual
psychotherapy for alcohol problem treatment. However, controlled studies have found that
individuals diagnosed with Alzheimer’s Disease may be successfully treated for
depression (Teri, 1994; Teri, Logsdon, Uomoto & McCurry, 1997). Results of these
studies, which employed cognitive behavioral techniques, tend to support the use of
specifically tailored forms of psychotherapy for individuals with other disorders.
Modifications to treatment for those with mild cognitive impairment include a slower pace
of therapy, greater repetition of material covered in therapy, heavier weighting of
behavioral as opposed to cognitive techniques, and involvement of family members in
treatment (Knight, 1996; Teri & Gallagher-Thompson, 1991). The presence of cognitive
impairment in the treatment sample indicates that these relatively simple modifications to
therapy should be considered in order to maximize treatment effectiveness. For those with
greater impairment, additional outreach effort and further modification of treatment may
be necessary.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
92
In summary, the proposed developmental distress model incorporated factors that
proved promising in explaining treatment entry for older male veterans. Even in a smaller
than expected sample (n = 70), the regression analysis used to test the model explained
significantly more of the variance in treatment entry than previous formulations have
found (Krause, 1990). The additional factors included in this model, psychological distress
and cognitive status, have particular relevance to older adults, especially to the low-
income male population among whom alcohol problems are prevalent and severe.
Measures of depression and anxiety did not have the significance expected, while
cognitive status emerged as relatively more important. This suggests that future
formulations of the model may need to lay greater emphasis on the role of impairment and
relatively less emphasis on the role of psychological distress.
Additional Clinical Implications
In addition to those recommendations described above, results of the study have
more general implications for alcohol problem screening and treatment in an older male
population. In terms of screening, results suggest particular characteristics of an individual
most likely to respond to an offer of alcohol treatment: younger in age, not married, better
educated, higher in substance abuse severity, higher in symptoms of depression, and
higher in cognitive functioning. Because rates of treatment entry tend to be low relative to
the number of individuals with a substance abuse problem, it is important to maximize the
effectiveness of case-finding outreach efforts (Stephan et al., 1992). Awareness of factors
correlated with treatment seeking may better enable mental health professionals to target
those individuals likely to respond to an offer of treatment, such as older adults in the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
93
younger age range (the “young-old”), who are unmarried, with depression, higher
cognitive functioning and relatively severe alcohol abuse. Results may also help to
identify bamers to treatment entry for those less likely to seek help, such as older adults in
the older age range (the “old-old”), who are married, with relatively lower CAGE scores,
lower levels of depression and lower cognitive functioning.
Study results regarding demographic correlations of treatment seeking behavior
have implications for how treatment offers are tailored to older adults. Most significantly,
the current study found that age was negatively associated with treatment interest and with
participation in a pre-treatment evaluation, independent of the effects of severity and other
demographic factors. As described above, cognitive functioning may play a role in this age
effect at treatment entry. At the stage of screening, cognitive factors might also be
expected to come into play, possibly explaining the age effect.
There may be additional age-related barriers to treatment seeking for older adults,
the cause of which has yet to be identified. Several hypotheses regarding this problem may
be drawn from the literature on mental health service use by older adults. Clinical reports
have pointed out that reluctance to seek substance abuse treatment may be based on
misunderstandings that older adults have about mental health treatment in general. Older
adults may lack basic information about mental health treatment, and may associate
therapy with inpatient care for psychotic individuals (Knight, 1996). Tactful education of
patients in the course of screening procedures could help reduce this ignorance of
treatment procedures. Older adults may be particularly sensitive to the stigma associated
with substance abuse, perhaps because alcohol consumption was viewed as a crime and a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
94
moral weakness during the Prohibition period (Osgood, Wood, & Parham, 1995).
Therefore, it is possible that couching language regarding both the screening procedures
and offers of treatment in terms of medical health, rather than mental health or addiction,
may be more effective with older adults than a direct offer of substance abuse treatment
(Blow, 1998). Such strategies may help to counteract possible cohort-linked hesitation of
older adults to seek treatment.
Limitations of the Study
The sample in this study was drawn from a population in particularly great need of
substance abuse treatment-medically ill, low-income, elderly male veterans. While the
study addressed important issues regarding this specific population, it may be incorrect to
generalize results to the elderly population at large.
The design of the present study focused exclusively on patient variables, to the
exclusion of variables related to the interviewers and hospital staff who interacted with the
patients. This limitation applies primarily to stages 2 and 3, in which variability in the way
patients were treated was likely. For example, the study did not measure the amount of
follow-up that individuals were given between screening, evaluation, and treatment.
Results of the study assume that all patients were treated equally, but there is no way to
know if this was the case since the extent of staff contact with the patients was not
measured. Because this study examined a population with multiple psychological and
social problems, including depression, cognitive impairment, transportation difficulty and
homelessness, it seems particularly likely that extent of staff efforts to encourage
participation in evaluation and treatment would have some influence over the response of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
95
the patients. Within the behavioral model, these may be important facilitating factors for
substance abuse treatment entry. Future studies should take staff and service variables into
consideration when examining predictors of treatment seeking behavior.
As noted above, the third phase of the study was also hampered by inconsistent
collection of data regarding anxiety, depression, and mental status of individuals evaluated
for treatment. Most of the individuals evaluated completed only some of these measures,
not all of them. This suggests the possibility of inconsistent treatment of individuals in the
study in ways that may have altered the stage 3 results. In addition, the inconsistent
administration of these measures resulted in a smaller than expected sample size for the
stage 3 regression analysis. This reduced statistical power, and limited the ability of the
investigator to examine the relationship between the variables in the proposed model.
While the stage 1 and 2 analyses had adequate statistical power, associated with large
sample size, the stage 3 model could not adequately be tested due to reduced power, with
the result that the three regression analyses could not be compared in regard to their
relative merit in explaining treatment seeking behavior. For example, there was a
significant mean difference in education between those who enrolled in treatment and
those who did not enroll at stage 3 (for this mean comparison, total n = 250). However,
due to missing data, many cases had to be dropped from the regression analysis, and the
effect of age may have been eliminated due to small sample size, rather than because of
relatively low importance of education relative to other variables in predicting treatment
entry (for the regression analysis, n = 70). In future studies, administration of all measures
at a single evaluation session would reduce the ability of patients to drop out mid-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
96
evaluation, providing a better opportunity to examine how scale scores predict treatment
entry.
The path models also illustrate differences in statistical power across all three
stages of the analysis. For example, certain standardized regression coefficients in the
stage 3 model were non-significant even though they appeared substantial in terms of their
size, e.g. the CAGE to depression path coefficient of .23, and the age to MMSE path
coefficient o f-.22 were both non-significant (see Figure 10). In contrast, relatively small
path coefficients stage 1 were significant, due to much larger sample size, e.g., the path
coefficient of marital status to drug use of .09 (see Figure 8). This contrast suggests that
the validity of the stage 3 model relative to those of stages 1 and 2 could not adequately be
measured in the present analysis.
Avenues for Future Investigation
Differences in findings regarding prediction of treatment interest versus treatment
entry suggest that future investigation of treatment seeking models must pay greater
attention to factors that distinguish these different dependent variables. For example, the
current study found that drug use was associated with treatment interest but not later
stages of treatment seeking, whereas CAGE score was associated with both outcomes.
Likewise, being married had a significant, independent effect on treatment interest at stage
1, but no such effect was observed on patient behavior at stage 2. Such discrepancies
suggest that treatment seeking models may benefit from greater precision with regard to
the specific outcome being tested. In other words, not all treatment seeking behaviors may
equally affected by a given variable. Future investigations should develop greater
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
97
theoretical and empirical linkages from these independent variables to the dependent
variables that they supposedly influence.
Explaining the influence of age on treatment seeking is an important area of future
study. As noted above, the negative association of age with treatment seeking behavior
was not entirely explained by the factors examined in the proposed model: greater age was
positively associated with being married, having lower education, lower severity and drug
use-all factors that were negatively associated with treatment interest. Yet when these
factors were controlled, the effect of age was still significant at stage 1 and 2 (Figures 8
and 9), but not at stage 3. These findings indicate that some other factors associated with
age, such as cognitive status, may contribute to the lower likelihood of older adults to seek
treatment. Cohort-linked attitudes regarding substance abuse treatment may need further
study. If cohort-based stigma around mental health in general, and around alcohol
treatment in particular is operating, this might help to explain the negative association of
age with treatment seeking. Future studies may wish to examine empirically the attitudes
towards having an alcohol problem and stigma around therapy that older adults have, to
see if such attitudes play a role in treatment seeking behavior.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
98
Conclusion
This study examined the predictors of substance abuse treatment seeking behavior
in a sample of elderly male veterans with multiple social and psychiatric problems, based
on a modified formulation of the behavioral health services utilization model. Regression
analyses found that younger age, being single, CAGE score, and drug use predicted
treatment interest. Younger age and CAGE score predicted participation in a pre-treatment
evaluation. For both these dependent variables, the effect of age was only partially
mediated by the effect of CAGE score, and was significant independent of other variables
in the model. This finding indicated that older age had a negative effect on treatment
seeking that was not explained by other variables in the model. In the final stage of the
study, only CAGE score and cognitive functioning predicted actual treatment entry.
CAGE score emerged as the most significant of the predictors, across three stages of the
data analysis, demonstrating the primary importance of need in the models proposed.
Results of the study have the potential to better explain what influences and enables older
adults to enter alcohol problem treatment, and to create more accurate models of mental
health treatment seeking behavior. Screening and outreach programs should take these
findings into account, to improve access to substance abuse treatment for older adults.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
99
References
Allison, P. D. (1999). Multiple regression: A primer. Thousand Oaks, CA: Pine
Forge Press.
Andersen, R., & Newman, J. P. (1973). Societal and individual determinants of
medical care utilization in the US. Milbank Memorial Fund Quarterly. 51. 95-124.
Atkinson, R., Tolson, R. L., & Turner, J. A. (1990). Late versus early onset
problem drinking in older men. Alcoholism: Clinical and experimental research. 14. 574-
579.
Barlow, D. FI. (1988). Anxiety and its disorders. New York: Guilford.
Barron, R. M., & Kenny, D. A., (1986). The moderator-mediator variable
distinction in social psychological research: Conceptual, strategic, and statistical
considerations. Journal of Personality and Social Psychology. 51. 1173-1182.
Beck, J. G., Stanley, M. A., & Zebb, B. J. (1999). Effectiveness of the
Flamilton Anxiety Scale with older generalized anxiety disorder patients. Journal of
Clinical Geropsvchologv. 5. 281-290.
Blow, F. C., Consensus Committee Chair. (1998). Substance abuse among older
adults. Rockville, MD: U.S. Department of Health and Human Services.
Blow, F. C., Brower, K. J., Schulenberg, J. E., Demo-Dananberg, L. M., Young, J.
S., & Beresford, T. P. (1992). The Michigan Alcoholism Screening Test— Geriatric
Version (MAST-G): A new elderly specific screening instrument. Alcoholism: Clinical
and Experimental Research. 16. 372.
Blow, F. C., Cook, C. A. L., Booth, B. M., Falcon, S. P., & Friedman, M. J.
(1992). Age-related psychiatric comorbidities and level of functioning in alcoholic
veterans seeking outpatient treatment. Hospital and Community Psychiatry. 43. 992-995.
Brennan, P. L., & Moos, R. H. (1991). Functioning, life context and help-seeking
among late-onset problem drinkers: Comparisons with nonproblem and early-onset
problem drinkers. British Journal of Addiction. 86. 1139-1150.
Buchsbaum, D. G., Buchanan, R. G., Welsh, J., Centor, R. M., & Schnoll, S. H.
(1992). Screening for drinking disorders in the elderly using the CAGE questionnaire.
Journal of the American Geriatrics Society. 40. 662-665.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
Bucholz, K. K., Sheline, Y. I., & Helzer, J. E. (1995). The epidemiology of
Alcohol use, problems, and dependence in elders: A review. In T. Beresford & E.
Gomberg (Eds.k Alcohol and aging (pp. 19-41). New York: Oxford University Press.
Chan, A. W. K., Pristach, E. A., & Welte, J. W. (1991). Detection of alcohol
problems and heavy drinking in three populations (Abstract). Alcoholism: Clinical and
Experimental Research. 15. 375.
Cohen, J., & Cohen, P. (1975). Applied multiple regression/correlation analysis for
the behavioral sciences. New York: Lawrence Erlbaum Associates.
Cohen, S. & Wills, T. (1985). Stress, social support, and the buffering hypothesis.
Psychological Bulletin. 98. 310-357.
Curtis, J. R., Geller, G., Stokes, E. J., Levine, D. M., & Moore, R. D. (1989).
Characteristics, diagnosis, and treatment of alcoholism in elderly patients. Journal of the
American Geriatrics Society. 37. 310-316.
Dunham, R. G. (1981). Aging and changing patterns of alcohol use. Journal of
Psvchoactive Drugs. 13. 143-151.
Ewing, J. (1984). Detecting alcoholism: The CAGE questionnaire. Journal of the
American Medical Association. 252. 1905-1907.
Finlay son, R. E., Hurt, R. D., Davis, L. J., & Morse, R. M. (1988). Alcoholism in
elderly persons: A study of the psychiatric and psychosocial features of 216 inpatients.
Mavo Clinic Proceedings. 63. 761-768.
Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). Mini-mental state: A
practical method for grading the cognitive status of patients for the clinician. Journal of
Psychiatric Research. 12. 189-198.
Gambert, S. R., & Katsoyannis, K. K. (1995). Alcohol-related medical disorders of
older heavy drinkers. In E. Gomberg & T. Beresford (Eds.), Alcohol and aging (pp. 70-
81). New York: Oxford.
Gomberg, E. S. L. (1975). Prevalence of alcoholism among ward patients in a
veterans administration hospital. Journal of Studies on Alcohol. 36. 1458-1467.
Gomberg, E. S. L. (1980). Drinking and problem drinking among the elderly.
Alcohol, drugs and aging. Ann Arbor, MI: University of Michigan Institute of
Gerontology.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
101
Gomberg, E. S. L. (1995). Older alcoholics: Entry into treatment. In T. Beresford
& E. S. L. Gomberg (Eds.), Alcohol and aging (pp. 169-185). New York: Oxford
University Press.
Gomberg, E. S. L ., & Nelson, B. W. (1995). Black and White older men: Alcohol
use and abuse. In E. Gomberg & T. Beresford (Eds.), Alcohol and aging (pp. 307-323).
New York: Oxford University Press.
Hamilton, M. (1959). The assessment of anxiety states by rating. British Journal of
Medical Psychology. 32. 50-55.
Hamilton, M. (1960). A rating scale for depression. Journal of Neurology.
Neurosurgery, and Psychiatry. 23. 56-62.
Helzer, J. E., Carey, K. E., & Miller, R. H. (1986). Predictors and correlates of
recovery in older versus younger alcoholics (1986). In G. Maddox, L. N. Robins, & N.
Rosenberg (Eds.), Nature and extent of alcohol problems among the elderly (pp. 83-99).
New York: Springer.
Hingson, R., Scotch, N., Day, N., & Culbert, A. (1980). Recognizing and seeking
help for drinking problems: A study in the Boston metropolitan area. Journal of Studies on
Alcohol. 41. 1102-1117.
Hingson, R., Mangione, T, Meyers, A. & Scotch, N. (1982). Seeking help for
drinking problems: A study in the Boston metropolitan area. Journal of Studies on
Alcohol. 43. 273-288.
Jellinek, E. M. (1960). The disease concept of alcoholism. New Haven, CT:
College and University Press.
Kasl, S. V., Ostfeld, A. M., Berkman, L. F., & Jacobs, S. C. (1987). Stress and
alcohol consumption: The role of selected social and environmental factors. In E. Gottheil,
K. A. Druley, S. Pashko, & S. P. Weinstein (Eds.), Stress and addiction (pp. 40-60). New
York: Brunner Mazel.
Kilpatrick, D. G., Roitzsch, J. C., Best, C. L., McAlhany, D. A., Sturgis, E. T., &
Miller, W. C. (1978). Treatment goal preference and problem perception of chronic
alcoholics: Behavioral and personality correlates. Addictive Behaviors. 3. 107-116.
Knight, B. G. (1996). Psychotherapy with older adults (2n d Ed.). Thousand Oaks,
CA: SAGE Publications.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
102
Krause, N. (1987). Chronic financial strain, social support and depressive
symptoms among older adults. Psychology and Aging. 2. 185-192.
Krause, N. (1990). Illness behavior in later life. In R. H. Binstock & L. K. George
(Eds.), Handbook of aging and the social sciences (3rd ed., pp. 227-244). New York:
Academic Press.
Krause, N., Dowler, D., Liang, J., Gu, S., Yatomi, N., & Chuang, Y. L. (1995).
Sex, marital status, and psychological distress in later life: A comparative analysis.
Archives of Gerontology and Geriatrics. 21. 127-146.
Magaziner, J. S., Bassett, S. S., & Hebei, J. R.(1987). Predicting performance on
the Mini-Mental State Examination. Journal of the American Geriatrics Society. 35. 996-
1000.
Miller, W.R., Brown, J. M., Simpson, T. L., Handmaker, N. S., Bien, T. H.,
Luckie, L. P., Montgomery, H. A., Hester, R. K., & Tonigan, J. S. (1995). What works? A
methodological analysis of the alcohol treatment outcome literature. In R. K. Hester & W.
M. Miller (Eds.) Handbook of Alcoholism Treatment Approaches (pp. 12-44). Boston:
Allyn and Bacon.
Miller, W. R. (1985). Motivation for treatment: A review with special emphasis on
alcoholism. Psychological Bulletin. 98. 84-107.
Moos, R. H., Mertens, J. R., & Brennan, P. L. (1993). Patterns of diagnosis and
treatment among late-middle-aged and older substance abuse patients. Journal of Studies
on Alcohol. 54. 479-487.
Moras, K., DiNardo, P. A., & Barlow, D. H. (1992). Distinguishing anxiety and
depression: Reexamination of the reconstructed Hamilton Scales. Psychological
Assessment. 4. 224-227.
Mueller, T. I., Lavori, P. W., Keller, M. B., Swartz, A. Warshaw, M., Hasin, D.,
Coryell, W., Endicott, J., Rice, J., & Akiskall, H. (1994). Prognostic effect of the variable
course of alcoholism on the 10 year course of depression. American Journal of Psychiatry.
151.701-706.
Osgood, N., Wood, H. E., & Parham, I. A. (1995). Alcoholism and aging: An
annotated bibliography and review. Westport, CT: Greenwood.
Pfeiffer, W., Feuerlein, W., & Brenk-Schulte, E. (1991). The motivation of alcohol
dependents to undergo treatment. Drug and Alcohol Dependence. 29. 87-95.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
103
Prochaska, J. O., DiClemente, C. C., & Norcross, J. C. (1992). In search of how
people change. American Psychologist. 47. 1102-1114.
Ritchie, K., Kilea, D., & Robine, J. M. (1992). The relationship between age and
the prevalence of senile dementia: A meta-analysis of recent data. International Journal of
Epidemiology. 21. 763-769.
Riskind, J. R., Beck, A. T., Brown, G., & Steer, R. A. (1987). Taking the measure
of anxiety and depression: Validity of the reconstructed Hamilton Scales. Journal of
Nervous and Mental Disease. 175. 474-479.
Schaie, K. W. (1996). Intellectual development in adulthood. In J. E. Birren & K.
W. Schaie (Eds.) Handbook of the psychology of aging (4th ed., pp. 266-286). New York:
Academic Press.
Schonfeld, L., & Dupree, L. W. (1991). Antecedents of drinking for early- and
late-onset elderly alcohol abusers. Journal of Studies on Alcohol. 52. 587-592.
Schuckit, M. A., Atkinson, J. H., Miller, P. L., & Berman, J. (1980). A three year
follow-up of elderly alcoholics. Journal of Clinical Psychiatry. 41. 412-416.
Schuckit, M. A., & Pastor, P. A. (1979). Alcohol-related psychopathology in the
aged. In O. J. Kaplan (Ed.), Psychopathology of aging (pp. 211-227). New York:
Academic Press.
Scogin, F. R. (1994). Assessment of depression in older adults: A guide for
practitioners. In M. Storandt & G. R. VandenBos (Eds.) Neuropsychological assessment
of dementia and depression in older adults: A clinician’s guide (pp. 61-80). Washington,
DC: American Psychological Association.
Sobel, M. E., (1982). Asymptomatic confidence intervals for indirect effects in
structural equations models. In S. Leinhart (Ed.), Sociological methodology 1982 (pp.
290-312). San Francisco: Jossey-Bass.
Stall, R. (1987). Research issues concerning alcohol consumption among aging
populations. Drug and Alcohol Dependence. 19131. 195-213.
Stephan, M., Swindle, R. W., & Moos, R. H. (1992). Alcohol screening in the
Department of Veterans Affairs medical centers. In R. E. Parry (Ed.), Screening for
alcoholism in the Department of Veterans Affairs. Washington, DC: Department of
Veterans Affairs.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
104
Tabisz, E., Badger, M., Meatherall, R. Jacyk, W. R., Fuchs, D., & Grymonpre, R.
(1991). Identification of chemical abuse in the elderly admitted to emergency. Clinical
Gerontologist. 11. 27-38.
Tarter, R. E. (1995). Cognition, Aging and Alcohol. In E. Gomberg & T. Beresford
(Eds.), Alcohol and aging (pp. 82-98). New York: Oxford University Press.
Teri, L. (1994). Behavioral treatment of depression in patients with dementia.
Alzheimer4 s Disease and Associated Disorders. 8 66-74.
Teri, L., & Gallagher-Thompson, D. (1991). Cognitive-behavioral interventions for
treatment of depression in Alzheimer’s patients. The Gerontologist. 31, 413-416.
Teri, L., Logsdon, R. G., Uomoto, J., & McCurry, S. M. (1997). Behavioral
treatment of depression in dementia patients: A controlled clinical trial. Journal of
Gerontology: Psychological Sciences. 52B. P159-P166.
Timm, N. H.(1970). The estimation of variance-covariance and correlation
matrices from incomplete data. Psvchometrika. 35. 417-438.
Tombaugh, T. N., McDowell, I., Krisjansson, B., & Hubley, A. M. (1996). Mini-
Mental State Examination (MMSE) and the Modified MMSE (3MS): A psychometric
comparison and normative data. Psychological Assessment. 8 .48-59.
Tucker, J. A., Vuchinich, R. E., Harris, C. V., Gavomik, M. G., & Rudd, E. J.
(1991). Agreement between subject and collateral verbal reports of alcohol consumption in
older adults. Journal of Studies on Alcohol. 52.48-155.
U.S. National Institute on Alcohol Abuse and Alcoholism (1979). Alcohol and
Health. Third special report to the Congress. (PHEW Pub. No. ADM 79-832).
Washington, DC: U.S. Government Printing Office.
Velasquez, M. M., Carbonari, J. P., & DiClemente, C. C. (1999). Psychiatric
severity and behavior change in alcoholism: the relation of the transtheoretical model
variables to psychiatric distress in dually diagnosed patients. Addictive Behaviors 24.
481-496.
Weisner, C. (1993). Toward an alcohol treatment entry model: A comparison of
problem drinkers in the general population and in treatment. Alcoholism: Clinical and
Experimental Research. 17. 746-752.
White, R. E., Luckie, L. F., & Miller, W. R. (1996). Suppose they offered and
alcohol abuse intervention, and no one came. Federal Practitioner. 51-57.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
105
Woodruff, R. (1973). Alcoholics who see a psychiatrist compared with those who
do not. Quarterly Journal of Studies on Alcohol. 34. 1162-1171.
Vogel-Sprott, M. & Barrett, P. (1984). Age, drinking habits, and the effects of
alcohol. Journal of Studies on Alcohol. 45. 517-521.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Alcohol expectancies and consumption: Age and sex differences

PDF

Family functioning, intergenerational conflict, and psychological symptomology of Asian American college students

PDF

Integration of science and practice: A collective case study of scientist -practitioner programs in counseling psychology

PDF

History of depression, antidepressant treatment, and other psychiatric illness as risk factors for Alzheimer's disease in a twin sample

PDF

Development and validation of the Cooper Quality of Imagery Scale: A measure of vividness of sporting mental imagery

PDF

Associations and mechanisms among attention deficit hyperactivity disorder symptoms, cognitive functioning, and drinking habits

PDF

A cluster analysis of the differences between expert and novice counselors based on real time training

PDF

An investigation of a new diagnostic sub-type: Post traumatic stress disorder with psychotic features

PDF

Inaccuracies in expert self -report: Errors in the description of strategies for designing psychology experiments

PDF

Dementia caregiving and ethnicity: African American caregivers and the sociocultural stress and coping model

PDF

Interhemispheric interaction in bilateral redundancy gain: Effects of physical similarity

PDF

Collective efficacy, anxiety, creativity/innovation and work performance at the team level

PDF

A context for timing, conditioning and modification of cerebellar function

PDF

Attachment style, interpersonal guilt, parental alcoholism, parental divorce and eating disordered symptomatology in college women

PDF

Factors that influence students' investment of mental effort in academic tasks: A validation and exploratory study

PDF

Content analysis of articulated thoughts of chronic worriers

PDF

An integrative analysis of racism and quality of life: A comparison of multidimensional moderators in an ethnically diverse sample

PDF

Depression in offspring following severe prenatal stress

PDF

A two-dimensional model of cognitive empathy: An empirical study

PDF

Determining competent/resilient outcomes across multiple domains in sexually abused and nonabused females

Asset Metadata

Creator Satre, Derek Davies (author)

Core Title Alcohol treatment entry and refusal in a sample of older veterans

School Graduate School

Degree Doctor of Philosophy

Degree Program Psychology

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

Tag OAI-PMH Harvest,psychology, clinical

Language English

Contributor Digitized by ProQuest (provenance)

Advisor [illegible] (committee chair), [illegible] (committee member), Walsh, David A. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-108783

Unique identifier UC11338024

Identifier 3027775.pdf (filename),usctheses-c16-108783 (legacy record id)

Legacy Identifier 3027775.pdf

Dmrecord 108783

Document Type Dissertation

Rights Satre, Derek Davies

Type texts

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

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

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