LIBRARY
Michigan State
University

 

 

 

PLACE N RETURN BOXtomnovothi-thflomyum

TO AVOID FINES Mum on Of Mon dd. duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Mandi" Action/Emil Opponmity lnstiMion
mm:

______________

A LONGITUDINAL STUDY OF THE PERSONAL NETWORKS OF INDIVIDUALS
RECEIVING TREATMENT FOR MENTAL ILLNESS
AND SUBSTANCE ABUSE PROBLEMS
By

Kurt Morgan Ribisl

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Psychology

1995

ABSTRACT
A LONGITUDINAL STUDY OF THE PERSONAL NETWORKS OF INDIVIDUALS
RECEIVING TREATMENT FOR MENTAL ILLNESS
AND SUBSTANCE ABUSE PROBLEMS
BY

Kurt Morgan Ribisl

There has been increasing recognition that social forces, such as social support and social
networks, influence substance abuse treatment outcome. Although individuals with
psychiatric disorders are fairly prevalent among individuals receiving substance abuse
treatment, the literature on the social networks of individuals with these dual disorders is
practically nonexistent. Therefore, the present study utilized a structured social network
methodology to study the social networks of 467 dually diagnosed individuals.
Participants were interviewed upon admission to a public psychiatric hospital and then 2-
and 6-months post-discharge. A model was proposed that predicted both relapse and
problem alcohol or drug use following treatment based on selected pre- and post-
discharge variables. Hierarchical multiple regression and logistic regression were
employed to test this model and to assess the impact of social network/support variables
and the role of psychiatric distress in predicting treatment outcomes.

The results indicated that poorer outcomes were consistently associated with

having contact with substance using social network members and experiencing greater

psychiatric distress. Other variables related to positive treatment outcomes, albeit less
consistently, were employment status, longer length of treatment, having family members
involved in treatment, and having more confidants or intimate network members. This
study supports prior empirical studies showing that friends, family, and other network
members can either promote or undermine a person's recovery from substance abuse
problems. Finally, for the first time in a dual diagnosis population, this study
demonstrates that severity of psychiatric problems was the strongest predictor of

substance abuse treatment outcome.

Copyright by
KURT MORGAN RIBISL
I995

ACKNOWLEDGEMENTS

A project as complicated as a dissertation requires a small cast of supporting
individuals. I have been most fortunate to be blessed with the companionship and
assistance of my wife Erin. Not only did she employ numerous tactics to help me remain
sane throughout this project, she also tirelessly helped me enter and double-check the
numbers in the Results section. Certainly, only true love can explain her assisting me in
such essential, albeit tedious, tasks. Thanks for the advice, comedic diversions, and help
as I struggled through this project...and probably most important, thanks for putting up
with me.

My best friend Paul Krecke stimulated me with his support and curiosity about
how science works. I will always fondly remember those important discussions and
debates about life that have occurred on the ultimate field and atop El Azteco. And
Dave, thanks for plying me with beverages at Paul Revere's Tavern the night before my
defense. This thoughtful distraction prevented me from worrying too much. Finally,
thanks to Gary Meyer, I was able to brainstorm dissertation ideas and exercise all at the
same time. The only other non-academic folks I would like to acknowledge are two
females who were there for me throughout the entire study. Although I was often

envious of the carefree lives you both lead and the frequent napping you did while I

toiled with analyses, to Nike and Schmu - thanks for keeping my lap warm while I typed
these pages.

My dissertation committee was truly top-notch. Thanks for inspiring me to do the
best possible job on this project. William Davidson, my chair, did a great job of shaping
my ideas into a manageable study. Neal Schmitt's guidance on the analyses was
invaluable. Finally, thanks to Tom Reischl and Bertram Stofi‘elmayr for their substantive
experience in this topic area and their encouraging words along the way.

In closing, I wish to acknowledge my student colleagues in the Eco program. I
especially wish to recognize Maureen Rumptz, Barbara Schillo, and Madeline Wordes.
Although our dissertation support group meetings were short-lived, our friendships have
not been. Also, Bonnie BootsMiller has been a fantastic co-worker and a true friend. I
look forward to keeping in touch with all of you as many of us migrate from East Lansing

to other destinations.

vi

TABLE OF CONTENTS

 

LIST OF TABLES .................................................... x
LIST OF FIGURES ................................................... xi
INTRODUCTION .................................................... 1
Purpose ....................................................... 4
Description of the Literature Search .................................. 5
Relapse Following Substance Abuse Treatment ......................... 7
Frequency of Relapse ....................................... 7
Contributors to Relapse and Successful Remission ................. 8
Summary ............................................... 10

Social Network Concepts ......................................... l 1
Definition ............................................... 11

Types of Social Networks ................................... l 1
Measurement of Personal Networks ........................... 12

Social Networks and Health Behaviors ......................... 14
Summary ............................................... 15

Social Factors and Substance Abuse Treatment Outcome ................. 16
Social Support and Treatment Outcome ........................ 22

Social Networks and Treatment Outcome ....................... 23
Studies Incorporating Both Social Support and Social Networks ..... 24
Methodological Shortcomings of Prior Studies ................... 25
Sample size and composition ........................... 25

Measurement ...................................... 30

Lack of longitudinal design or adequate follow-up periods . . . . 30

High attrition ...................................... 30

Summary ............................................... 3i
Psychiatric and Substance Abuse Comorbidity ......................... 32
Prevalence of Dual Disorders ................................ 32
Treatment Difficulties Encountered with Dual Diagnosis ............ 33
Hypotheses of the Present Study and Rationale for Model ................ 34
Pre-Discharge Variables .................................... 3 5
Post-Discharge Variables ................................... 36

vii

METHOD .......................................................... 4O

Participants ................................................... 40

Design ....................................................... 44

The Experimental Treatment Program ............................... 46

The Usual Care Program ......................................... 46

Community F ollow-up Procedures .................................. 47

Measures ..................................................... 49

Addiction Severity Index ................................... 49

Symptom Checklist -lO .................................... 53

Social Network and Sobriety Support .......................... 54

Family Involvement in Treatment ............................. 55

Demographics ........................................... 55

Data Analysis .................................................. 55

RESULTS .......................................................... 57

Descriptive Characteristics of Social Networks ........................ 57

Missing Data Analyses ........................................... 61

Intercorrelations Among the Study Variables .......................... 64

Regression of Substance Use Outcomes on Study Van'ables ............... 68

Hierarchical multiple regression .............................. 69

Logistic regression results ................................... 73

Direct and Indirect Pathways Leading to Substance Use .................. 79

DISCUSSION ....................................................... 92

Network Characteristics .......................................... 92

Role of Pre-Discharge Factors ..................................... 93

Social Networks, Sobriety Support, and Psychiatric Distress .............. 95

Limitations ................................................... 99

Future Research ............................................... 10]

Conclusions .................................................. l 02
APPENDIX A

SAMI Project Participant Consent Form ............................ 105
APPENDIX B

Social Network Measure ........................................ 108

APPENDIX C - l
Standardized Beta Coefficients for the Path Model Predicting Addiction
Severity Index Alcohol Composite Scores ........................... l 10

viii

APPENDIX C - 2
Standardized Beta Coefficients for the Path Model Predicting Addiction
Severity Index Drug Composite Scores ............................. 111

APPENDIX C - 3

Partial Correlations for the Path Model Predicting Alcohol Relapse ........ 112
APPENDIX C - 4

Partial Correlations for the Path Model Predicting Drug Relapse .......... 113
LIST OF REFERENCES ............................................. 1 14

10.
ll.
12.
l3.
14.

IS.
16.

LIST OF TABLES

Background information on research reports relating social support and

social network variables to substance use or treatment outcomes ....... 17
Methodological shortcomings of research reports relating social support

and social network variables to substance use or treatment outcomes . . . . 27
Demographic Characteristics of Dual Diagnosis Participants at Baseline . . 43
Design of the Dual Diagnosis Study ............................. 45
Measures and Time Periods for the Current Study .................. 51
Descriptive Statistics on Social Network Characteristics at 2-month

(Time2) and 6-month (Time3) Follow-up ......................... 59
Comparison Among Participants Having Complete and Missing Data on
Demographic and Substance Use Variables ........................ 62
Zero-Order Correlations of All Demographic Variables, Predictor

Variables, and Outcome Variables in the Hierarchical Regressions ...... 66
Hierarchical Multiple Regressions of A81 Alcohol and Drug Composite
Scores on Demographic and Social Network Variables ............... 71
Logistic Regressions of Alcohol and Drug Relapse on Demographic

and Social Network Variables .................................. 76
Direct, Indirect, and Total Effects on ASI Substance Use Outcomes ..... 87
Direct, Indirect, and Total Effects on Relapse Outcomes ............. 9O
Standardized Beta Coefficients for the Path Model Predicting Addiction
Severity Index Alcohol Composite Scores ....................... 1 IO
Standardized Beta Coeflicients for the Path Model Predicting Addiction
Severity Index Drug Composite Scores ......................... 111
Partial Correlations for the Path Model Predicting Alcohol Relapse . . . . 112
Partial Correlations for the Path Model Predicting Drug Relapse ...... l 13

SAPPJNr-r

LIST OF FIGURES

Hypothesized panel model predicting substance use ................. 39
Trimmed panel model predicting ASI Alcohol Composite Score ........ 83
Trimmed panel model predicting ASI Drug Composite Score .......... 84
Trimmed panel model predicting alcohol relapse .................... 85
Trimmed panel model predicting drug relapse ...................... 86

xi

CHAPTER I
INTRODUCTION

The abuse of alcohol and other drugs has created a significant health problem in
the United States. Approximately 100,000 deaths each year are attributable to misuse of
alcohol and 20,000 deaths to illicit drug use (McGinnis & Foege, 1993). Alcohol was
estimated to be involved in over 39% of all fatalities from traflic crashes in 1991,
contributing to over 16,000 traffic deaths (Zobeck, Stinson, Grant, & Bertoluci, 1993).
Moreover, alcohol-related traffic crashes differentially impact younger persons; alcohol-
related traffic crashes alone were implicated in 49 percent of years of potential life lost
for males and 34 percent for women in 1991 (Zobeck, Stinson, Grant, & Bertoluci, 1993).

Some of the key indicators of our nation's substance abuse problem have shown
that positive changes are occurring; however, for many of the other indicators no change
has occurred or the problems are getting worse. For instance, although deaths from
traffic crashes have been declining in recent years, substantial increases have been
observed for drug-related deaths due to AIDS cases that occur among injection drug
users or from people who have had sexual contact with them (Institute for Health Policy,
1993). Although national surveys have shown encouraging declines in the numbers of
indmduals reporting using cocaine, consumption of cocaine has merely leveled off while
the number of heavy cocaine users has continued to increase over the last decade

(Everingham & Rydell, 1994).

2

The negative consequences of substance abuse, however, extend beyond
mortality. The economic impact of substance abuse is substantial. Drug abuse in 1990
cost the nation approximately 70 billion dollars; the economic cost of alcohol abuse is
even higher, and has been estimated at over 98 billion dollars (Institute for Health Policy,
1993). In addition, substance abuse can also place severe strains upon families (e. g.
Holden, Brown, & Mott, 1988 ); approximately one in four Americans report that alcohol
has been a cause of trouble in their family (Institute for Health Policy, 1993).

Breaking the grip of addiction is not an easy process. Quitting or becoming
abstinent is usually attainable; however, maintaining abstinence is typically the
challenging part of successfirl recovery (Marlatt, 1985b). This is why Mark Twain is
frequently quoted as saying that quitting smoking is easy-~because he had done it
hundreds of times. The pioneering work of Marlatt and Gordon (1985) has greatly
improved our understanding of the relapse process, how to prevent relapse, and how to
prevent a lapse or slip from becoming a fiJll-blown relapse. This line of research has
suggested that most relapse episodes have causes that may be internal (e.g. cravings,
mood states) and/or external (e. g. direct and indirect social pressure) (Daley & Marlatt,
1992; Marlatt, Baer, Donovan, & Kivlahan, 1988).

There has been increasing interest in studying the social factors related to both
relapse and prolonged abstinence. Because factors outside of treatment probably have as
much of an impact on recovery as the treatment experience (Billings & Moos, 1983;
Westermeyer, 1989), several researchers have sought to explore how social influences,
such as social support and social networks, can promote positive outcomes. In a general

sense, social support refers to actions that are perceived as supportive or helpfiJl,

3

whereas, the social network is simply the connections among individuals, some of which
may provide support (Ritter, 1988; Thoits, 1982).

For individuals who already have alcohol or drug problems, the social network
may help influence the type of alcoholism treatment sought (Strug & Hyman, 1981) and
whether the individual subsequently is able to maintain abstinence from the problem
substance. For instance, among individuals who have successfully modified their
problem drinking behavior, many report that one of the most important factors in staying
sober was the support they received from their family, friends, and partner (Sobell,
Sobell, Toneatto, & Leo, 1993; Tuchfeld, 1981). However, network members can also
facilitate relapse. For example, the Narcotics Anonymous Handbook tells readers that
"old fi’iends, places and ideas are oflen a threat to our recovery. We need to change our
playmates, playgrounds and playthings" (Narcotics Anonymous, 1988, p.15).

Although prior studies have explored how social networks influence the initiation
of alcohol and other drug use (Ellickson & Hays, 1992; Newcomb & Bentler, 1988) and
how the existing personal network can afl‘ect substance abuse treatment outcomes
(Goehl, Nunes, Quitkin, & Hilton, 1993; Hawkins & Fraser, 1987), a thorough literature
search failed to retrieve any published studies of the social networks of individuals with
substance abuse problems and a co-existing psychiatric diagnosis (i.e. the dually
diagnosed). This apparent void in the substance abuse and dual diagnosis literatures is
vexing given that psychiatric problems tend to co-occur with substance abuse problems at
an alarming rate. For example, estimates of the proportion of individuals in substance
abuse treatment who have a psychiatric diagnosis range from 20 to 65 percent, depending

on the population and study methods (El-Guebaly, 1990; Ross, Glaser, & Germanson,

4

1988). In addition to being fairly prevalent, having psychiatric problems may be related
to having a poorer prognosis. Dually diagnosed patients have higher rates of
hospitalization, homelessness, and violent and criminal behavior than patients with either
psychiatric or substance use problems (Osher & Kofoed, 1989). Also, the level of co-
existing psychiatric problems was the strongest predictor of poor treatment outcome in a
study of over 800 persons with alcohol and other drug problems (McLellan, Luborsky,
Woody, & Druley, 1983 ).

Given the multitude of problems facing dual diagnosis clients and the extensive
demands that they place upon service delivery systems (Bachrach, 1986-1987), there is a
pressing need to learn more about how to intervene with this population. A social
network approach holds promise. A person's social contacts, their peers and friends, can
influence an individual to remain drug-free or can facilitate their relapse (Brown, Vik, &
Creamer, 1989). The pressure from network members to use substances can be great and
can create what has been termed a "magnetic" pull back into addiction (Stephens &

_ Cottrell, 1972). Understanding how the social network affects persons with dual
diagnoses can be instrumental in shaping intervention strategies and providing a better
understanding of the web of social influences that may contribute to their sobriety or
relapse.

Bumps:

The literature relating social support and social networks to substance abuse
treatment outcome has indicated the potential importance of these variables in
understanding relapse and the successful resolution of alcohol and other drug problems

(Fraser & Hawkins, 1984a; Westerrneyer, 1989, for example). However, most of these

5

studies are plagued with serious methodological shortcomings, such as small sample
sizes, high attrition, and poor measurement of social network variables. Furthermore, the
number of individuals receiving substance abuse treatment who have co-occurring
psychiatric problems is rather high, but the literature on the social networks of
individuals with both of these dual disorders is practically nonexistent. The purpose of
the current study, therefore, is to examine how psychiatric distress, social support, and
social networks are related to short-terrn relapse and substance abuse problems among a
dual diagnosis sample that had recently received inpatient treatment.

Most of the methodological shortcomings of prior studies have been addressed in
the current study. In the current study, the sample size is large (over 300), the design is
prospective and longitudinal, and a standardized methodology was employed for
assessing the social network. Before the literature review is presented, a brief description
of the procedures for conducting the literature search is provided.

Computerized literature searches using PsycINFO (1976-1994, Spring) and
Sociofile (1/74 - 4/94) supplied the majority of the references for the literature review.
Key words for searching were: Drug use/usage, alcohol use, substance abuse, addiction,
mental disorders, mental illness, mental patients, and social networks. A boolean search
was conducted on each database separately (e. g. social networks AND {mental illness
AND substance use}). No citations were found on either database, therefore, the search
was modified slightly to include articles on social networks and either mental illness or
substance use. This revised search yielded 294 English language citations in PsychINFO

and 38 citations in Sociofile. Another search was attempted by using the keyword dual

6

diagnosis, which netted 337 PsychINFO citations, but only 1 article dealing with social
networks. The keyword dual diagnosis is not used by Sociofile.

Additional criteria for inclusion in the present literature review were the
following: (a) the articles must have been written in English; (b) the study population
needed to include adults, therefore articles on the networks of children and adolescents
were excluded; (c) only articles that focused primarily on social networks were included;
(d) articles on individuals with the dual diagnoses of mental retardation and substance
abuse were excluded because these individuals did not meet the eligibility criteria of the
treatment wards used for recruiting the sample of the present study; (e) articles focusing
on the HIV/AIDS risk behaviors of the social networks of injection drug users were
beyond the scope of this topic. Additional references obtained from the reference
sections of the articles from the original search were also used in this review.

This review starts by providing an overview of the intrapersonal and interpersonal
factors that influence both relapse and recovery from alcohol and other drug problems.
Next, this review addresses definitional and measurement issues in the social support and
social network literature and discusses the potential mechanisms whereby social
networks may influence substance use. A critical examination of the specific literature
on social support and social network influences upon substance abuse treatment outcome
is then provided. Special attention is given to the methodological quality of these studies.
The final section focuses on the t0pic of dual diagnosis, especially upon the prevalence of
co-occurring disorders and how psychiatric problems are related to poorer treatment

prognosis.

7
WWW

One of the general conclusions from the substance abuse treatment outcome
literature is that treatment typically has a positive, although quite modest effect and that
most individuals who eventually relapse do so shortly after treatment completion
(Longabaugh & Lewis, 1988). This section examines the frequency of relapse following
substance abuse treatment and then explores the reasons for relapse or successful
abstinence.

Ezequemflfielarzss:

Although relapse is operationalized differently across studies, most studies have
defined relapse simply as re-use of the drug (Stephens & Cottrell, 1972). In some
instances, multiple categories reflecting a continuum of subsequent use are created
instead of a simple binomial relapse category. For example, in a random sample of 922
individuals who had made contact with alcoholism treatment centers, Polich, Armor, and
Braiker (1980) split their 4-year follow-up sample into individuals who were: Drinking
with problems (54%), drinking without problems (18%), and abstinent six months before
follow-up (28%). Although it does not appear very encouraging that 72 percent of the
participants had consumed alcohol prior to their six month interview, the authors caution
that this reflected a substantial improvement from baseline when 90 percent of the cohort
was experiencing serious drinking problems.

In a follow-up study of 200 individuals addicted to narcotics (Stephens &
Cottrell, 1972), 87 percent had re-used narcotics within six months of hospital discharge.
When this relapsing group was subdivided, 10.5 percent had used narcotics only a few

times, 12 percent used narcotics regularly without becoming re-addicted, and 64.5

8
percent were re-addicted. In a study that combined results from over 600 patients from
three different treatment centers, it was found that only 10 percent of alcoholics had
maintained continued abstinence 18 or 24 months later (Gottheil, Thornton, Skoloda, &
Alterman, I982). Wallace's study of cocaine smokers revealed that 76 percent relapsed
in three months and 94 percent in six months.

All of these more recent findings are congruent with an earlier review of 84
studies examining relapse for various addictions (e. g. heroin, cigarette smoking, and
alcohol addiction) that found that about two-thirds of relapses occur within three months
and over 90% occur within six months following treatment (Hunt, Barnett, & Branch,
1971). There is also a striking similarity in the patterns of relapse for each of these
addictions when these are graphed (although alcohol is slightly different, having a lower
relapse rate). In summary, the relapse rates following treatment appear to be quite high,
but more importantly, what are the factors contributing to the return to drug use or to the
maintenance of sobriety?

Marlatt and colleagues (1985c) have developed a useful and widely used scheme
for categorizing relapse episodes. Two primary determinants of relapse have been
proposed, factors that are intrapersonal (including reactions to environmental events) and
interpersonal. The intrapersonal categories include: (1) Coping with negative emotional
states, (2) Coping with negative physical-physiological states (including withdrawal and
cravings due to withdrawal), (3) Enhancement of positive emotional states, (4) Testing
personal control, and (5) Giving in to temptations or urges. The interpersonal

determinants are composed of: (1) Coping with interpersonal conflict, (2) Social

9
pressure, and (3) Enhancement of positive emotional states. The social pressure
component includes both direct pressure, such as provision of the substance and verbal
pressure, or also indirect pressure, which includes modeling of substance use to the user.
Interestingly, it appears that the type of social pressure may interact with the type of
substance. For example, Marlatt suggested that most drinking relapses due to social
pressure are caused by direct social pressure, but indirect mechanisms (i.e. modeling) are
more salient for smoking.

Wallace (1989) examined the psychological and environment determinants of
relapse for 35 crack cocaine smokers. Although the coding scheme for that study
differed slightly from Marlatt's categorization, it was found that most relapse episodes
(86 percent) were caused by multiple factors. The most popular psychological
determinants were a relapse in response to a painfiJl state and failure to enter arranged
aftercare treatment. If relapses caused solely by either psychological or environmental
determinants were considered, more relapses were caused by environmental factors (17
percent versus 9 percent). The most common environmental contributor (N = 12) was the
environmental stimuli of people, places, and dmgs. Lack of family support or being
homeless was implicated in five episodes. In a study of 200 males addicted to narcotics,
similar factors influenced relapse, such as the use of substances for coping with social
problems, depression, and the influence of addict friends and their environment (Stephens
& Cottrell, 1972).

In studies of individuals whose alcohol or drug problems have gone into
remission. both intrapersonal and interpersonal factors appear to promote recovery. To

understand the processes of spontaneous remission in alcoholics, Tuchfeld (1981)

lO

conducted in-depth interviews with 35 men and 16- women who had resolved their
drinking problems on their own. The primary reasons for stopping drinking included
various health concerns, exposure to educational materials about alcoholism, religious
involvement, interventions by families and friends, financial or legal problems, and
extraordinary events (e. g. personal humiliation, attempted suicide, exposure to negative
role models). Many of these factors appear to be related to the presence of social support
from family, friends, and other network members. Tuchfeld noted that among people
reporting that loved ones influenced their recovery, many had persistent support from a
spouse or family member. In a more recent study of 182 individuals who reported having
a current alcohol problem or who had successfully recovered without treatment, similar
factors (e. g. support from spouses, friends, and family) emerged as being important to
recovery (Sobell, Sobell, Toneatto, & Leo, 1993). In fact, spousal support was endorsed
by the greatest number of participants (two-thirds) as being the most helpful factor in
maintaining the resolution of alcohol problems. Moreover, participants in the study by
Stephens and Cottrell (1972), reported that support of the family (27 percent) and support
of non-addict friends (14 percent) were important reasons for abstinence.
Summary

The literature on the relapse process suggests that relapse is rather common after
alcohol or drug treatment, nonetheless, individuals do show positive improvements in
controlling their substance use. Relapses tend to occur shortly after treatment and most
relapses occur within six months following treatment completion. Relapses have both
intrapersonal and interpersonal origins. Furthermore, individuals who have successfully

resolved their substance use problems frequently report that support from family and

ll

friends was instrumental in their recovery. The next section provides an introduction to
the social network concepts because many studies of the social influences upon recovery
have utilized social network techniques.
SociaLNetmrkLanems

Definition

The terms ”social network" and "social support" have been used quite loosely in
the literature and although these terms have been used interchangeably, they are distinct
concepts (Thoits, 1982). The most widely cited definition of social networks is that they
are "a specific set of linkages among a defined set of persons, with the additional
property that the characteristics of these linkages as a whole be used to interpret the
social behavior of the person involved" (Mitchell, 1969, p. 2). In contrast, social support
refers to actions that are perceived as being helpful or supportive. Therefore, the focus of
social networks is human interaction and social contacts, which may or may not be
indicative of the amount of social support available (Fraser & Hawkins, 1984b; Gottlieb,
1985; Israel, 1983). One of the major advantages of employing a social network
approach over examining only social support, is that a network approach is more neutral
allowing for the investigation of both supportive and unsupportive relationships and it
permits the examination of numerous network characteristics beyond social support that
may contribute to health status (Israel, 1983).
ImutLSnciaLhLetmrks

Generally, there is a distinction made between open and closed social networks
(Mueller, 1980). The unit of study in open networks, which are usually called personal

or egocentric networks, is a focal individual and their relations to other members. When

12

the entire population of interest and their interactions are studied, the network is said to
be closed. This is usually only possible for smaller, clearly bounded populations due to
the amount of effort and cooperation required of respondents and also the massive
amount of data that can be generated (Laumann & Schumm, March 1992). An advantage
of the personal network approach is its ability to be used for studying large populations
with many network variables (Mueller, 1980). F urtherrnore, personal networks permit
the use of standard multivariate statistical techniques for data analysis (Laumann &
Schumm, March 1992). Most studies in the area of mental health, physical health, and
drug use have utilized the personal network approach.
Measummenmfflctsnnamstmrks

The personal network approach was taken in the present study and, therefore, is
of greater interest than the closed network. The most common method of assessing
personal networks is first to enumerate the network members (Marsden, 1990). This is
usually done by asking the focal individual (also called the ego or star) several name-
generating questions. For instance, one of the most common questions asks the
individual to name the individual(s) whom they would turn to if they needed to talk about
an important personal problem. This approach is based on an exchange theory of social
relations, which posits that people who are sources of rewarding social interactions are
most likely to shape the focal person's attitudes, behavior, and well-being (McCallister &
Fischer, 1983).

The number of name generators used to enumerate the network may affect the
resulting network size. The appropriateness of different numbers of network delineation

questions was examined in a study of the general population in the Netherlands using a

13

name generator approach to enumerating the network (Van der Poel, 1993). All
participants were asked ten probe questions (e. g. Who would you talk to if you are
depressed? Who would take care of you if you were sick for a couple of days? With
whom do you have contact at least once a month?) When subsets of the 10 probe
questions were varied to determine the minimum number of questions necessary to
describe the network, it was concluded that three is the absolute minimum number of
questions to obtain some indication of the network size, but five name generators yielded
considerable improvements. Three questions explained only about 55% to 60% of total
network size determined from 10 questions, but 5 questions explained about 84% of the
total network size.

Once the network has been enumerated, detailed questions called name
interpreters can then be asked about each of the network members (or alters) (Laumann
& Schumm, March 1992). For example, it is common to inquire about the basic
demographic characteristics of the alters (e. g. sex, age, race/ethnicity, relationship) and
the type and frequency of interactions. A matrix that features the names of the alters is
sometimes constructed, which permits the investigator to assess systematically whether
each alter knows every other alter. Sometimes, however, this procedure for establishing
network density is only performed on a subset of alters. Further questions about the ego
and the alters are often developed based on the specific purposes of the study. For
instance, the name generating questions may be customized to the study population (Van
der Poel, 1993). Based on the detailed information collected from the alters, it is then

possible to calculate several qualitative and quantitative indices for the social networks.

For example, the fimctionaLcharamfistigs (e. g. affective or instrumental support) and

14

structural characteristics (e. g. size of network) can be assessed (Israel, 1983; Mitchell &
Trickett, 1980) and related to other variables of interest.
5.!” l llillEl'

Network members can help influence individuals to make positive health changes.
For instance, studies from the health promotion literature have shown that support fi'om
network members has been related to improved diet (Zimmerman & Connor, 1989) and
quitting smoking (Coppotelli & Orleans, 1985; Mermelstein, Lichtenstein, & McIntyre,
1983). Sometimes, support from network members is not related to smoking cessation,
but mere exposure to other smokers has been negatively related to abstinence (Curry,
Thompson, Sexton, & Omenn, 1989). Moreover, positive health changes in husbands
have been observed even when they were not the target of a dietary intervention, but their
spouses were (Shattuck, White, & Kristal, 1992).

Networks can also provide negative influences whereby network members may
introduce the use of cigarettes or illicit drugs to their friends or partners. For example,
among female inpatients at a drug treatment program, one-fifth reported that they were
initiated to crack cocaine through a family member and more than half through a male
friend (Boyd & Mieczkowski, 1990). Parental modeling, which is often confounded with
genetic factors, is also important. Comprehensive reviews by Robins and Smith (.1980)
and Hawkins, Catalano, and Miller (1992) have indicated that having a parent who uses
alcohol or drugs is consistently related to increased likelihood of use by their children. In
several instances, however, these reviews indicate that peer influences are stronger
predictors of the initiation of substance use, especially for females who are often

introduced to the substances by male partners. More recent studies of adolescents

15
confirm the finding that peer and family drug use is a risk factor for adolescent use
(Newcomb & Bentler, 1988) and that offers of alcohol or drugs from family, friends, and
other drug users was related to later drug use (Ellickson & Hays, 1992).

The mechanisms of how networks influence the health status and behavior of
their members are not well understood (Hammer, 1983; Ritter, 1988). Ritter has
examined some of the potential explanations. For instance, the networks may be useful
in providing information to their members, may provide care and specific aid leading to
improved health, or they may even provide a form of social control that influences
members by peer pressure. Others posit that the process can be explained by Social
Learning Theory and the importance of modeling (Akers, Krohn, Lanza-Kaduce, &
Radosevich, 1979). The effects of network may be explained by classical conditioning
and social learning, whereby, a former drug user may be exposed to drug-related cues
(e.g. addicted friends, a particular environment, or the substance itself) that may elicit
cravings or positive outcome expectancies about drug use (Goehl, Nunes, Quitkin, &
Hilton, 1993; Marlatt, 1985a).

Summary

In summary, social network members and the support they can provide can have
an important influence upon whether an individual is able to successfully adopt positive
health behaviors and modify their negative health behaviors. In some cases, perceived
social support is more strongly tied to behavior changes, whereas in other instances the
modeling of the target behaviors by network members appears to be a more salient factor.

In addition, intimate network members, such as a spouse or partner, seem to be the most

l6

important individuals affecting the successful maintenance of the desired behavior
change.
SQciaLEactnrmniSuhstanceAhuselrcammnflutcnme

Previous studies that have examined the role of social support or social network
influences upon substance abuse treatment outcomes are presented in Table 1. As
mentioned earlier, the distinction between social support and social networks has not
always been maintained in the literature. In some instances, when a study discussed clear
social network influences but labelled them social support, they are placed in the social
network column. The key differentiating factor was that when the independent variables
were based on the ties among individuals (both supportive and unsupportive), they were
placed in the network column. Some of these studies are cross-sectional, but most are
longitudinal and almost all have utilized treatment samples. They are organized in the
table in reverse chronological order by type of study--socia1 support only, social network

only, or both social support and social network variables assessed.

17

0928:. .a «axe—Eu .

SEE—vs gas—oz: .

#538: 2958 E3» vac—gov =03 .
5 E 525235 3E8... .

c286 .3393:
guinea... .a Baa—as .

0.5.22. :39; .268 tone—9.23 =03 .
g

 
 

352:: £23.25“
5332:: - tax—cam

38.2.23
8.23:.
53983.5:
.m> 59.395 .
3.3m
2.8 D 8.58%
o».2_35>5:=2um
2.253; .

€988.95
.=_3.mo..=um .mciouv

9.24.83
3294.293; 32%

mwau

Eu...) .5 mi»: 3.8M.
3225 ..o coaanfiia
20:83

m£>= 8: 3:5,.
.323: .3 95:36.52.
5 E 353.. 39.320
.3 5.39253
<2?.<

E 580% a mag:
<2)? ... 8.35:8

j. 3 x53. .8

 

_ ...2533 _

22.3555

29 5 2.2.33.5;
25 52255
3.3m 95:33—

.28» BREE
.coaazw 33m 1:280

.3263. 9.80
.336 E253:
3.3. 35.33

a 82.. 3:de

3:23 .3528. cc.

Azco flew

a 5-...é 562.93
.5595 C3 E253:
was a .282
$5.39: an

Ammo:

25.5.59 9% 3.3m

A53:

2.3.5.3 an 3510

 

 

:2... «NEE-"=5... A233 2
miractsaw . 3185:35390; .
395% 9:029:12 38 253502 mflfiladam 5.59
52. Lot 358 2.95%»: . .5253 .355 :3: . «43.5 a. MES“. 5.6.33 3235.5
adam— ant. ace—5:335 ca :5: «.3583. .
anus—uh... - amémm . 22:9 .oz . 2.....-mm3 p.253 35:9 9:38:32 am .
madman 3.1.5.3 .Azefi 3:35 . 28:9 33252 on . 3%: ..a .120
A8933 E2535 .3: 023.33 may. 2.52:0 83ata> 8323.; c255 .Euow $28932 Amy—053..
.oococmifi; 888:8 “SE8.“— 3 3.20.. 83at~> ..o 933:5 33:09.0 A 2% 3.83.02 item
guano
:.:.....=. . .. .. . ..... x... . ... ...a-o... .... m. ... ...... ...,..:...:o. .. -..I...“

. _ DEE.

l8

 

 

 

.3...» ~— .56

 

29.29. 39228.. . 8.5.6.3... .95.. 2.35.55 . 99:925.... 9,98 n .3 22.52.95. .
.8325. . :9... 352...... 985.80 : .. oz 5 ... 8.3.2:...
3.8.. 3.8.58 .3... 5.53 822...... . .9... ... N .... 2883.. .q u 8.. 8.3 seam .
v9.9.9.2... . on: new .2... . 2.98.99: 92.88.92 93...... .... 2.9.2330 . 3.8:
29.9.9.2... . 8.2.9.5050 . o. 298 299.3 9...... z. ... 8.3.9.5.. 8...... .... .9 8.3.:
3.5.3. 9.3.95.3: 2.0... 8.. 92.930 . .95.. 2.99:2?" .9525» 2925.3795993 mm .
2.2.. a. 2......
2. .9... 0. 9.3.2.0 .
5930.... 9.35....
wit... 99.69.
a. 2.0.2.999... .oZ .
.9425. ..o 9...... o
2.930....
9.32.5. mafia—9.29..
2.2... 5992.0 .
.52....
.30.... 592.00 . .....mm... «5......m
29.292... . .....wm. 3.2.2.9. 45... um... . 9.9.... 29.2992... .
29.9.9. Enigma: - 9.52.... 2.0.... 2.99.50 . 3...... ... ... 9m... . .....-mm... t9...=m
dug 9.3 .226? .0... o . ...—2...... @9393... . $3393-30... .
292:9. 2.9.22.3: . 9.3.2.... 25... 599200 . .53.. 3mm....... 3:32;”. 20.2.09 20 . $8..
$9.29. 39.22.92. . 93 .209... .o... o . ..moh 2.2.99.8 cont—.5 3.59.9992... 3.22.3.9 38.9.9391
.flw. 53.3.... 5.20.2 . 5...... 38m :56 . .o 5......O .3... .. . .1. 5.3520
29.9.9.
2.9323. - .2989... 55...? .0 995.53un . 2.9.5.0225 2H 8.
29.9.9. 29328.. - €5.25... «8......» =39>O . a. .35....1...=Ea .. A 5 99.33
29.9.... 39.423. . QEHEEF... 9.93 2.9.5.9924. . Can... ...... 29.97.25”. 9.9.93 :4. . €32.33 Ea... 32.3....
0.322299 .9 . _ o. 9.. 20.2.8299. 3:33... o 922...... 39.2.9990 .
V9.39. 29328.. 9.0.9.. 3.9—. .02 . + 9.33 :89... ...... mam . 22.53.35... .
- 22:... 2.2,... 9.29... 5.52.5 .... 995.3323. . 322.523. .....N. ...—:92 5...... ...... .23.}. 22.32.... .4. 5......557. . 32.:
32.7 .....igvaux . «5......1. .3930 ....9295 329.22.... 3 . .- .9 5.5:
.333. 2.9.2.3.. .3: 92.33.... 89. 98092.0 993.....a> 33......) «5.....5 229.5 $335592 3.55:4.
62.95%. 8632... 9323.. 0. 29.9.9. 8...a_.a> ... 939....../ 2.922930 Ava 4.93.92 32%

 

19

 

 

«9.2m

m 2.. 33.55:...
:23. 030.02m
30:50.5.-.» N ...oE
o .8. 753.38..

 

00.8200 30.53.33. p.850 E2542.
3.5.... 3.3:... .02. 0:05 6892000.. v:- n.0..=00 . 3:02.200 ..0 .02 . 3:33.92... ..0 5:33 Em._0..00.- 3.50.3.—
0005 59.028 8 «.0503 339...: 83.3.3 225.3300 ...—0.3 . a .a 95.303. .
an... 303...:00 80:. as. 000.» 089,000.: . ..0 089.002. .200. ..0 .02 . .&.au¢0..a.2 £05000 $5.553 m: .
.0005 N .850 o... .3... 9-30:0. ....» $3.00... 3003082... ..0 5:33 8:23.- 089? an .
£3.03 ....03 80:. VS. «.0503 3.5.8800 . m .a 3.2.. mix—ED . 1.0030 .02 . .0255...Eu 3.03... . “2.2.00... 3.8.608 an . 2mm. 3 ..a .u .002
E n a.
8 .... A C . 00:00.3. $83503
..0 .0012. .uowcox. . .00 .35.... 5.03.. w. 0. .25.“:
3.22:: . 83st? on: 35.33 :4 . on. 00.3.33 .m3 «000..» 320... 20.2. an 52.5.... .
agdlmP/JB AF. .0: 06.3....— . 330:... ..0 .02 . 300.83-320 .
832.... . 83.5, 3.. ”8%.... 85o =< . a. .3. ... 3.. Eu. .uz. 5.58 $8.02.. 8.3:
303 #9.... 0. 3.32 39......3: 4503.2. .693... . 3:233. 9...... .02 . 4.03.00 ...00. 3?... 35:33.2 .... .9 5.3.5233
mmflfl $4.2 3....qu . 32...... .3 .oz . 35:2. .2 .
£230.25“. ..0 £00..» 50.. ..0.. 02.530 3.0...
£3039. A2320: 0. was 9:3 5.. :09...» 620 39.250
3.9. £253.00 0. £0.30 an... 339... 08.. .95 .3... 3.3.5.035...
2.0.5.0: c. .3530 395.0 3.32. .550 .932 13.035 .
o - wanna—ox 00.500800
00...“: 20.»...80 . «we... 3.: 0. 0m . .5537. ”103.02 .
300.0...30 m:.oco=cc. £2.82: {03.2. .030...— . 8.03.... .5430 .0E.m
8.22 .8: 22% new ..45362 . .-. 5.2... 3.3.9.3 . Ea :
....3...a0.. . .08 . .u 4.53.0: a. E03 95. .039“. . Es... £90.39... .08 m 3.: 95. 94.03.92 . 3.2.... ...—.359»; an. . .53.. a. 3.450:
6%.. 33.5 Aqua.
00.20.05. .2255 3.0 3.04....
Sum. 35?. a 6.3 2. 9.3 .55
- 3.33 «aroma...» m .9303 .8. 63...»...
...Cofi... mix—Eu 6.002. 320.59.
.833 .9. .930. .3... ma... .3050..- .a.00. .35.... 35.335. .0:— ..
.....5 3000.232. n 2.. 00.. 0.002. On. 0. an 80.... 020.3. .2. 03:33... 33.03... 9.002. 55.3.0... .a:.v:..m:0.. .
.5000... 3.... 4.03.2. emf-2,4. £9.03 3.09.6.9... :. 00.30.0330 .md. ...:3505... .35 7.. 03:059.? .
.2535? .0 309.558... 2.. ... 35.0.9.3; 35.. 0. .0 032.50.. . ..0 030903 {03.92 .53 0.9.0500 3.... .23 3 03082:.
.0. .59... .0 9.5.... 983 2.. 9.03 «3.03.02 2:3... fizzm :3. n. ...—.0. 309......0. N... . 0% $53....
A8033 .0253: .93 02.333 39. 0809.00 833.32 830...; €2.95 .326 .mm0.0.5£u./. A3553.
.ou:o=..£<. 8600.00 u>...u00 0. 003.2 8.93:5 ..0 0.02.8.2 Eng-0000 AZm. 4.03.02 7.3m

 

20

 

.0.a.:2...:mv 30.50..

 

 

00.0.0.5. - 31.950500 33.30:»... H. ..0 03:09.3 .07. . 3.7.30.3.
3.22 38...? - 08.3.5.3 . 20.2. 3.2. .2: m ... 2.8.3510. 32...... 2.55.32 ...... 33.23 .
00.0.0.5. - 95.0.3.6 was—...... .0 3053.4. 50.9.0. ..0 .00» x. 0.5.09.3 ....30..0...0 0.9....» ah .
00.0.0. .20.»...«0a - man—...... .0320... 9.55.0 ..0 03:253.... .02 o .03.... ..0 .02 . 3:0..00... 0.050.. 3 . .50.. 22.0.0002
Anotowofio 0.... 3.030... 3.7.30.3. ....o
002.300.00.20. N a... .0. «.009... ....Ea... . 65.03... 10.03.30. .
03 ms... v.0...0w0.3 00:05.2... M. £08 3.50.0.3. .00.:00 . 0.9.... x... .
0. 00.0.0. 3033...... - .0303 9.0.0 02.0.0.0... . m0... ....3 .00... 08.... 8.6.2.3. 0>... . 5.020.. .30..
00.0.0.5. . 9.9.95 0.2.5... . 80.32.33... 03 man. gum—MUM 00.20.36.000 gm . ..0885 .M 00a
00.0.0. .20.»...80 . 0.2.0.... w...3-...0.0.. .030... .
0...:
ma... 0. 00.0.0.5. 0.0.5.... ..0 .0... Q 0.00.6 0.... .
flag ”c.3238
00... ms... .08-N d .. 0. 0.2.5... ..0 00:08.... .
00.0.0. 20.0.3?! 0.5.5.30. 0.2.5... S. 9.30: . 0.3.50... 4.03.0: .....sm .04.. «.09....
.08-. ... 8.. ms 80. 0. 00.0.0. 8.8.3 .300... ..0 0.... £0.03 . 00.4003. .0 .0304 . 9-30:0. .0...
...... .322: 5.3 35.58 a3. .3 ...: . ...: 03.. .35. : a. 8.. ......m ...»... :95... as... .. 3.3.08. .
a 0.... ..0 .0530... .96 9». 4.0.5.0.. ..0.0>O . 030...... 00.0.0800 . 300...... 0.0.0... 0.0.0.. 9. . .30.. «.0005.
00.0.0.5. . 300.000 .02 .
60.0.0.
30.1.39. - «E030... .0. 0. 0m 0. 0.000.. 0.02 .
00.0.0. 203:8; - 2m ... €0.33 03.080. . 0.00.000 Zm
00.0.0. ...0... v.95... 00.0.0805. .
303.70.. . ...080wca... use... 033.8800 . €0.00... .0. 0....0003..0~. .
00.0.0. 303.70.. - 9.0.5... 000...... . 0. 0m 2000 0.000.. .02 . ......mok. w. 80:23..
00.0.0. 20......0. . €009... 2.80... . 00.3....» 3.0028. 80.... .
fig ...}. .o .5280 . 5.5... 5.8a
2.50.00 0. 00.0.0.5. 3.30.23 2m # Em £030.... Zm 905.00... «.09.... 0:0... .0030001 . -..0..E.0£.0.0....0.. an .
.3. 0.... 3...: . -..0: .9 0.39.3. ... 508000 .02 . «.09....3 ...—.... 002000”. . 30......0. mm . .23.. 309.33.
.8000... 2.06.00... .3... 02.55... 80. 0800.00 8......a> 830...; 9.3.96 ....00m 30.000502 Any—9.3.1
62.00.31... 000.0800 0.5.8.. 0. 00.0.0. 00......0> .0 0......0> 2005000 .29 u10.302 7.00m

 

21

 

 

 
 

 

 

 

.00.: «- 00>0 0.000000 3.7.50.3.
m. 0.3.2.0200. ... .2000. .23.. 153.08. .
00.0.00 303.03: - ZmDm . 200000. 050.. 000.00.. ..0 2m :. 000.. 8000.0 Sum: 0.90.3 xh . Ann-00..
030.005. - 4mm. o :0.:0n.00n. - 3 ..0 0.... 00:83.6 0.2.0.... .303 .0005... 5:0...2. 0:32.08 :5 o ..a .0 ...-003
.02—.0300
00000.. 0. 00.0.00 3.: 54:00.00 0.00.3.0:
wit... ...... 00.00.05 2.2. .:¢0.-..:m.n n ..0 .. .
e00.0.00 ..0.»..0m0:- 00......» 30.040. 00...00.$-0..V .
00.0.005- 0???» 30.090» 0.00.0... . .00..:.00-00 5104:0000. 5.2.5.3000 .3007.
020.005. - .00....3 50.500 3.80... . $000000 .... Anna—003.00 30:00.. 60.320000 .0 ...—.0...
003.00.... - 0004:...00; .0: 30.00:. 0:: 0.80.0..00 50.03300 ..0 2.50.... 95:23.3. “8.00330 a o
00.. 2.3 $02.65.... 50000000.. ..0 3.80.. ..0 .02 . ...... 5:82.350 083. 300-20: 0.000805
080.00.... - 0004:0050 $52.0 0. 03.. >00.:0>:. 0.3.; 3032.0»... 80...: . 3.7.50.3.
00.. 2.3 w:...:..0.:.=. :0....:.m.00 ..0 3.8.... ..0 .02 . 0: 00.6005. £0.03. 3:00.202. :0 .0..:. 9.-.. ...:.0....w:0-. .
0020.00 203.60..- 000450 .552... ..:0:..£< . .3175... 00 .... :. m:.:.00 «000...: 00.30.0305. . 30.0.00:
-00 .0: 00.. 0...: 3.30.0.0: 802003-00 .0 .02 . 00:00» .- 3.0.0000. 0.030000 00.. €2.03. 00 10000.. . ......30 00 00:300.... ...-00..
...mm . :7: 00.6. 0.00:0 .000... 0 . 3.00.235. on. . ..EN .0 :00000
30000.: 0—08.30. .00... 00:03... 80. 2:00.30 830.00., 0030.03, cacam .a.00m ~»..-.0.000...0$. 7.09.34
.02—0:034. 00:50.00 0.»...000 0. 00.0.00 0030.00> 00 0.50.00> .5025 .29 ...—03.02 7.00m

 

22
SmiaLSupmmdlmtmenLQumme

Six of the 16 studies featured in Table 1 related social support measures to
substance abuse treatment outcomes. In four of the studies, general social support
(including both family support and total support) measures were used, but in two studies,
support was conceptualized as greater involvement of family or fiiends in treatment.
General social support (e. g. emotional support, tangible support, etc.) was related to
improved substance abuse outcomes in two studies (Booth, Russell, Soucek, & Laughlin,
1992; Grey, Osborn, & Reznikoff, 1986), showed mixed results in one (Ohannessian &
Hesselbrock, 1993), and was not predictive of outcome in another (Oyabu & Garland,
1987). In the study by Oyabu and Garland (1987), however, drug use was not assessed,
but treatment success was defined as better social and psychological well-being.
Attempting to explain why support was unrelated to outcomes, these authors speculated
that general social support measures may not be appropriate for a substance using
population because the support questions ask about behaviors that may not be helpful to
someone struggling with addiction problems. For example, the support items ask the
person if they have people who support their actions, however, those actions may be
destructive ones, such as excessive drug use.

When family and friends provided support by being involved in treatment, more
positive treatment outcomes were likely. For instance, Siddall and Conway (1988) found
that individuals who were discharged successfully from treatment had greater family
involvement. In addition, individuals whose significant other participated in treatment
were less likely at follow-up to be using cocaine as measured by urine screens (Higgins,

Budney, Bickel, & Badger, 1994).

23
Scciallfletmrksandlreatmemflmmme

Only 3 of the 16 studies investigated the impact of social networks upon treatment
outcome. In these studies, treatment outcomes were related to network size (F avazza &
Thompson, 1984), the number of intimate members and network breadth (Westermeyer
& Neider, 1988), and the network composition and attitudes (Hawkins & Fraser, 1987).

In the study by Favazza and Thompson (1984), relapse was associated with
having a smaller social network, but this was based on only 12 cases. In fact, individuals
with poor outcomes constituted a group of only two persons. In an investigation of 168
patients labelled with a diagnosis of substance abuse, having fewer intimate network
members was associated with: Higher symptom reports on the SCL-9O and the Beck
Depression Inventory, more observed psychopathology on the Brief Psychiatric Rating
Scale, and higher scores on a modified Michigan Alcohol-Drug Screening Test
(Westermeyer & Neider, 1988). This study, however, was cross-sectional. In summary,
the relationship between network size and outcome is unclear because these relationships
have not been tested in well controlled, methodologically sound studies.

The impact of fellow network members who use substances probably is
associated with poorer outcomes. For instance, a three-month follow-up study of 38
individuals who had received substance abuse treatment found that having more drug
using network members was related to an increased likelihood of relapse (Hawkins &
Fraser, 1987). These networks had members strongly influencing the respondent to use
drugs (22.6 percent), but also influenced them against drug use (33.8 percent).
Furthermore, it appeared that the social networks stayed about the same size over time,

but for many respondents there were positive changes in network composition. New

24

members appeared to have been included in the networks because there were fewer long-
terrn acquaintances and fewer drug users at follow-up.
SI'I 'EIS'IS lS'lll 1

Assessing both social network and social support constructs in the same study
may allow for a better understanding of the social processes that may impact relapse.

The last seven studies cited in Table 1 investigated both network and support constmcts.
Two of the studies that have measured both social support and social network constructs
have suggested that general support has positive effects (Macdonald, 1987; Rhoads,
1983), although support was unrelated to outcome in the investigation by Joe and
Simpson (1983).

One advantage of assessing both support and network constructs in the same
study is that the efi‘ects of each variable on substance use can be tested independently. In
several instances the findings for support and networks have differed. Among methadone
maintenance clients, social support was not related to drug use as measured by the
number of drug positive urines, but having one close significant other who used drugs
was related to drug use (Goehl, Nunes, Quitkin, & Hilton, 1993). The impact was
significant: 63 percent of individuals with at least one drug-using network member
tested positive, compared to only 36 percent of those with no drug-using significant
others. Supporting this finding was the observation that having contact with current drug
users was related to relapse in a six-year follow-up study of opiate-addicted persons (Joe
& Simpson, 1983). Interestingly, Rhoads (1983) observed that males who avoided
contact with former drug users had better outcomes, but for females it was social support

that was related to positive outcomes (less substance use).

25

Social support has been conceptualized in many different ways in these studies.
Sometimes support was assessed by scales measuring the quality of support, whereas in
other studies support was defined by the number of supportive individuals in the person's
social network. For instance, the number of individuals that a person could go to in
response to problems (intimate members), was related to positive outcomes in the
investigation by Rosenberg (1983). The mere number of social contacts (i.e. the size of
the network) however, was unrelated to substance use.

In contrast to general social support, specific support for maintaining sobriety has
not been widely studied, and may deserve greater attention. Sobriety support was
unrelated to outcome in one study (Macdonald, 1987) and was actually related to
negative outcomes in another (Gordon & Zrull, 1991). However, this negative effect was
caused by support provided primarily by individuals who were not sober themselves,
perhaps confounding support and network member drug use.

Illll'lSl . [13.5].

The studies cited previously have many notable methodological weaknesses.
These shortcomings are highlighted in Table 2 and are discussed in the following
sections.

Sample size and composition, Most of the reviewed studies had small sample
sizes (N's from 38 to 60), which limits their statistical power and generalizability. Five
studies had sample sizes from 12 to 59 persons, four had sample sizes 60 to 99 persons,
and three studies had sample sizes larger than 100 persons. In many cases, these sample
sizes were too small for the application of more SOphisticated multivariate analyses.

Furthermore, individuals with co-existing psychiatric problems were not explicitly

26

included in any of the reviewed studies. In fact, individuals with psychiatric problems
systematically were excluded from five of them (F avazza & Thompson, 1984; Goehl,
Nunes, Quitkin, & Hilton, 1993; Gordon & Zrull, 1991; Higgins, Budney, Bickel, &
Badger, 1994; Rosenberg, 1983). Investigations by Westermeyer (1988) and Rhoads
(1983) were the only published reports that investigated social networks and both mental
health and substance abuse variables. Neither population, however, had co-occurring

psychiatric problems or were labelled mentally ill.

27

$8:

 

3095.039:
35:03-3er S S t -. S -- t 5 m2 an 5735.20
32538 2:5: €250
5&3 do Sentence 23:82.. -- z .. .. s -- s s 2: a =€Em
on: mac ..0 55.qu 2:838 .0:
2.. .2822: use? 8:68..
0.3 I 8.5.0 3.5 ”2.323; 38:
o. 3.2.532 :o 35:01 S S t t S t : 5 5 ._a .0 58m
2.33afi
on as... 5.353 2 Sta 2::
2: 53a :32 5:3 25.585 1.0
we» ... 52m 3336 «3&3 33m
5:0.an0 E253: ban a:
456:3 55385 95:35 .0: 30 A23 3 v5.30
3: 3:233 0553.0 .2. ED 5 : -- t S -- : 5 mm a. .5ng
€253 Econ 9.332:
2 2:3 _ tun: 5:0 29:8
cums. a 5; 3.3.38 2. 2:05 3::
3.38 8.0: 55:. $3: @000 S S : t S vows—02m 5 3 co ... .o 5.5
$558.5...» 33282.92 :.s3._m$_ 3: 5.8a 88...; u3292 59:5 .2 was 6.7. .A 3:23.
.5 8: 35.3:m co van—om

3:33.“.-bzfiso

 

 

3235 ..0 2:350 @0803 _= 3.258 n =4
min—coo Econ ..mEQ «ma—Q ”.939? u 97..
38m>3.m33§£:m
5.8a ”.8388
...... . .... ....4... ...-... .. .. .4. ....u. A... .... ... .........)

.N 033.

28

tobacco ...—00.9:

 

 

0.5.. 2003 00000.0. .38 .0... .000» 2.3 3
. ..o 9.0 05 .0 ©0508 9132.0... 5 -. S S S -- -- .5 8 325.0002
.30..
8:5... ......m s -- s s s -- s .. a... 82.2.... .... 8.
00.00.052—
0.:0..0 03.2.0 ......o ...va ...:0
505.800.. 89:23 .0 .0252. 090...
0.9:: :08... 2...... S S .. S S -. S .- A... 3.3.. ...—0005.
.30 S
00:02.3..3. .- -- 5 5 .3 $030.07.”.— -- S cm 305.503.
:3 C
2.2.3.8.. .07.... oz .5 -- S 5 S -- : .5 0mm .... .0 30.1.
Sn... 2
.2233 .305 -- s S 5 - -- s S we. .30....0503
tho 3 002......
0.953. ....Em S - 5 .5 -- -- .5 : .3. fl 2.3 50..
2.5.5.0..
1203.0 5.3333005 30..
..o...:.... ..m... .30..
0.00. 00.3.3.» 02 602.59....
0.30 07.2.3 0.00.352 S .5 -- .8 - v0.0202m. l S N. a. 039:...—
.conE00 30.53 .0 3.0.00.2.
0... 60:30.... .0... 0:50.00 0.6.9...
.00....0: 0.3.. .93.. E08029»...
.=0E.00.. 32. ..a 0.5.5.508 3.2.. V
03.2... 2.. as: .2. ...c .. - s -- 5 83.3.0 5 -- R ... .9 2...»...
awaiaocopfi $0323.23: 0.3.3.354 0»: E0395 00:05.34... 2.5.5.02 €2.95 .2 .503 9.7.. .../.. A3553.
.0 0.... 00:33.5 .0 00.....01
.n0=00_00.m-z...=!5
30.0.... ..0 0:50.00 $389.4. ... 2.0.5:. .. .2
0.3.50 380. ..we: 0...: 02.3.4. . .3...

«:06... >35 .0 00:830m
5.8... 08:23....

 

29

 

A}:
2.... :30....00 3.3.]. 5 5 5 5 5 5 5 5 2."... 500.50 2:.

U00... 3:95:00

 

 

 

0:0 .. n. .0: @0502... :03 £30..
.0: .0050... 5.3.0530 «.9202 5 5 -. 5 5 0000.02... 5 .- .... .... .0 ...000
28..
0.2.3.5.. 8.2.. oz 5 -- 5 5 5 33.0%. 5 5 on. ...:N a. 80.00
0:53:25 723.82.»... 22.3.05... 3: 52%.. 8:803... .52»... :25... .2 we: 0.... .... 0:953.
.0 0m: 00:30....m .0 300.0%
..0:0:.00..........:§O
.00.0.:. ..0 0:50.00 .3300}. ... ...-2.0.: u ....
2.0509 1.2% ”max. .mEA. .2007. » 0.x.
300.. .......m .0 00:33.5

.632. 08:23....

 

30

Measurement. Concerning the measurement of the social networks, many studies
have used different methods of enumerating the network, which makes comparisons
across studies difficult. Also, some studies have used network enumeration methods
with questionable reliability and validity. For instance, in one study (Favazza &
Thompson, 1984) individuals could be considered social network members if they were
dead or the individual had completely lost contact with them. These problems just
highlight that more work needs to be done to develop and refine social network and
support measures relevant to psychiatric populations (Mueller, 1980) and for addicted
persons (Oyabu & Garland, 1987).

Lmnmmmmgmmmmmmpsm Four of the studies in
the table were cross-sectional or retrospective, but to examine how networks influence
substance use typically implies a longitudinal design. There is a need for more
longitudinal social network studies to investigate the ways in which networks change and
to determine causal relationships among network characteristics and other variables (Ell,
1984; Mueller, 1980). Furthermore, the time periods chosen should reflect the fact that
the most critical period for studying relapse is the first six months following treatment
when most relapse episodes occur. Several studies had initial follow-up periods that
extended well beyond this time frame.

Highmmflnn. Several of the reviewed studies had substantial attn'tion, which
limits their external validity. F avazza and Thompson (1984) interviewed 42 participants
at baseline, but only located 14 and interviewed 12 participants (74 percent attrition) 11
months later. Joe and Simpson (1983) located 77 percent of their original target sample

and interviewed 66 percent, which is fairly respectable given a six-year follow-up period.

31

High attrition rates, however, are not unique to social network studies of substance
abusers. A meta-analysis of attrition rates for 85 substance abuse prevention studies
found that the average follow-up interview completion rates were: 81 percent at three
months and 78 percent at six months (Hansen, Tobler, & Graham, 1990).

High attrition, besides limiting statistical power due to a smaller sample size, can
also threaten the generalizability of the study if attrition is differential. This is fairly
common. Individuals reporting greater drug use at baseline have been more prone to be
missing from the follow-up sample (Josephson & Rosen, 1978) or are more difficult to
locate than other participants (Moos & Bliss, 1978). Rhoads (1983), for instance,
reported that the 49 respondents who had completed all follow-up interviews represented
only 38 percent of eligible participants. Furthermore, those lost to attrition tended to be
older, had a longer history of drug use, and were involved in more illegal activities.

Reviewing the literature on the relationship of social networks to physical health,
Berkman (1986) recommended that investigators expend the effort to minimize attrition
because several prior longitudinal network studies experienced fairly substantial attrition.
Moreover, in at least one of these studies, those who were socially isolated were more
likely to be unlocated at follow-up. Consequently, when this type of differential attrition
occurs, the generalizability of the findings and the ability to detect important substantive
relationships among variables can become compromised.

Summary

After treatment, positive network changes may occur such as a shift in social

network composition. It appears that successful maintenance of sobriety may be more

strongly influenced by substance-using network members than having support for

32

maintaining sobriety, but few studies have tested both factors simultaneously.
Unfortunately, many of the reviewed studies had serious methodological shortcomings
such as a small sample size, poor measurement of social network variables, or high
attrition. None of the reviewed studies were conducted with a dual diagnosis population
despite the increasing awareness of the prevalence of co-existing psychological disorders
in substance abuse populations.

ErexalenneofllluaLDisoLders

Psychiatric disorders and substance abuse disorders ofien co-occur for both
treatment populations and the general population. Among cocaine addicts seeking
substance abuse treatment, 55.7 percent met criteria for a current psychiatric diagnosis
and 73.5 percent for a lifetime diagnosis (Ziedonis, Rayford, Bryant, & Rounsaville,
1994). For 501 patients seeking assistance for alcohol or drug treatment, 65 percent had
a current Diagnostic Interview Survey mental disorder and 78 percent had a lifetime
mental disorder, in addition to their substance use (Ross, Glaser, & Germanson, 1988).
As cited earlier, prevalence estimates of substance abuse in psychiatric populations is
accordingly quite high, with prevalence estimates of 20 to 75 percent of patients
(El-Guebaly, 1990).

It is not appropriate to generalize the high rates of any co-morbid conditions that
occur among treatment populations to the general population because individuals with
several types of disorders are overrepresented in treatment settings (Kessler & Price,
1993). Incorrect generalization from treatment prevalence to true community prevalence

is known to epidemiologists as Berkson's bias. However, fairly high rates of co-existing

33

substance and mental illness problems also have been observed in non-institutionalized
populations. In the general population survey conducted for the National Institute of
Mental Health, the Epidemiological Catchment Area study, more than half of the
individuals with a substance abuse disorder qualified for a mental disorder; individuals
with a mental disorder were nearly 3 times more likely to have an addictive disorder
(Regier, Farmer, & Rae, 1990).

I [I'm 1' E I 'IIJ IE' .

Individuals with these dual disorders present several challenges to the treatment
system. For example, dually diagnosed patients: (1) use disproportionately greater
amounts of medical, social, and legal services; (2) require more staff time in psychiatric
and substance abuse treatment programs; (3) are more likely to disrupt medical and
nonmedical staff relations; and, (4) have extremely high recidivism rates (National
Institute on Drug Abuse, 1991). Indeed, in a study of 460 male alcoholic and 282 drug-
addicted persons from six different rehabilitation programs, the severity of the psychiatric
condition was the best predictor of response to drug abuse treatment (McLellan,
Luborsky, Woody, & Druley, 1983). Individuals with lowest levels of psychiatric
severity improved in every treatment program, but practically none of the individuals
with the highest levels of psychiatric severity showed improvement in any of the
treatment programs. In addition, a ten year follow-up study of former drug abuse patients
found poorer employment and psychiatric outcomes for individuals with chronic
psychosis compared to patients with acute psychosis or no evidence of psychosis

(Perkins, Simpson, & Tsuang, 1986).

34

Given the widespread scope of dual diagnosis and the treatment difficulties facing
individuals with these dual problems, promising approaches to treatment are urgently
needed. Unfortunately, there is a dearth of longitudinal data on this population that might
indicate which factors affect the long-term course of recovery (Brown, Ridgely, Pepper,
Levine, & Ryglewicz, 1989). Two longitudinal studies have been recently published, but
were based on preliminary reports each consisting of fewer than 30 participants
(Hoffman, DiRito, & McGill, 1993; Ries & Ellingson, 1990).

Further contributing to the lack of an adequate knowledge base on dual diagnosis
is the exclusion criteria employed in substance abuse and mental health studies. As
demonstrated earlier, investigations in the substance abuse literature commonly exclude
individuals with known psychiatric problems, which is also true of the psychiatric
literature whereby persons with substance abuse problems are similarly excluded from
study participation (Cohen & Kochanowicz, 1989; Wasylenki, Goering, Lemire, Lindsey,
& Lancee, 1993, for example).

As described at the end of Table 2, the current study remedies most of the
methodological shortcomings of prior social network studies on the topic of substance
abuse. The current study has a large sample size (n over 300), is prospective and
longitudinal, and was characterized by low attrition. Furthermore, a standardized social
network methodology was employed in the present study.

WWW

Based on the reviewed literature, a longitudinal model predicting alcohol and

other drug use following treatment is proposed and featured in Figure 1. Two general

classes of variables are used to predict substance use following discharge from treatment.

35

Pre-discharge variables include experimental condition, treatment duration,
demographics, and family involvement in treatment. Post-discharge variables include
sobriety support of network members, network member substance use, the number of
intimate network members, and psychiatric distress. Each of the variables in the model
and the hypothesized paths were carefully chosen based on the findings of prior empirical
studies. The rationale for the model is summarized below.

The direct effect of experimental condition upon reducing substance use is being
tested because numerous studies have shown that substance abuse treatment is effective
in decreasing later substance use and problems resulting from substance use (Institute for
Health Policy, 1993; Rydell & Everingham, 1994; Simpson & Sells, 1983). An
underlying reason for the effectiveness of treatment might be that the treatment program
(a) encourages the client to avoid former drug-using network members and (b) helps the
client muster support from network members for staying clean and sober. For example,
prior studies have demonstrated that individuals who showed the most improvement in
terms of reduced substance use successfiilly avoided their former drug-using social
contacts and increased the amount of social support they received (Hawkins & Fraser,
1987; Rhoads, 1983). Therefore, these paths are tested in the model. The path leading
from experimental condition to decreased psychiatric distress was not hypothesized
because both treatment groups in the present study (described later in the Methods
section) received similar psychiatric treatment; however, the experimental group was

given more intensive substance abuse services.

36

Another factor related to substance use outcomes is the duration of treatment.
Goehl and colleagues (1993) found that individuals who spent more time in treatment
were less likely at follow-up to have used drugs based on urinalysis results. Furthermore,
in an overview of the research contributions of the Drug Abuse Reporting Program,
Simpson and Sells (1983) also noted that longer treatment duration was related to greater
treatment benefit.

Several studies reviewed earlier showed that family support or significant other
participation in treatment has been related to greater retention in treatment (Siddall &
Conway, 1988) and more positive treatment outcomes (Grey, Osborn, & Reznikoff,
1986; Higgins, Budney, Bickel, & Badger, 1994). Therefore, the extent of family
involvement in treatment has been incorporated into the model. Demographic variables
have been included in the model as control variables because prior work has shown that
demographic characteristics such as sex, race, age, education level, employment status,
and marital status have been related to substance use or treatment outcomes (Lex, 1991;
Macdonald, 1987; Smith, 1993).

Given the problems with using general social support measures for addicted
populations mentioned by Oyabu and Garland (1987), a promising approach might be to
assess the specific support for maintaining sobriety that is provided by network members.
Although this construct has not been widely tested in the substance abuse field, support
for maintaining behavior change has been related to increased likelihood of quitting
smoking (Coppotelli & Orleans, 1985; Mermelstein, Lichtenstein, & McIntyre, 1983) and

for other healthy behavior changes (Zimmerman & Connor, 1989).

37

Network member drug use has been shown to adversely affect an individual's
sobriety in many studies. Individuals spending] more time with other drug users are
more likely to use opiates themselves (Hawkins & Fraser, 1987; Joe & Simpson, 1983;
Rhoads, 1983). The number of intimate network members, or confidants, has also been
related to substance use. Rhoads (1983) found that individuals lacking intimate
relationships reported greater drug use at follow-up. Among 168 persons with substance
abuse diagnoses, those with larger networks had fewer alcohol-related problems as
measured by a modified MAST (Westermeyer & Neider, 1988). Finally, successfully
recovered alcoholics reported more confidants than relapsed alcoholics two years after
study recruitment (Moos, Finney, & Chan, 1981). Thus, the number of close fiiends, or
the intimate network, has been related to substance abuse measures in several diverse
studies.

It was fithher hypothesized that affiliating with network members who use
alcohol or drugs will be more strongly related to relapse or problem substance use than
support fi'om family during treatment, sobriety support from the network, and the number
of intimate network members. This hypothesis was based on findings from the substance
abuse literature that have never been applied to a dual diagnosis sample. In the studies
reviewed in Table l, the adverse impact of substance using network members has been
related to relapse more consistently than the presence of the various forms of social
support.

The final post-discharge factor related to substance use was psychiatric distress.
The relationship between the severity of psychiatric problems and their negative impact

on treatment outcome has been demonstrated in research by McLellan and colleagues

38

(McLellan, Luborsky, Woody, & Druley, 1983), although their sample was not dually
diagnosed. Therefore, psychiatric distress is proposed as a predictor variable and not an
outcome variable because the role of psychiatric distress in predicting substance abuse
outcomes among a dual diagnosis population was not tested in any of the reviewed
studies.

One possibility not tested in any of the reviewed studies is that a person's
substance use may influence them to assemble a network of fellow substance users. In
other words, it is not the network that influences the substance use of the focal person,
but it is an active process of the focal person to structure their network based on their
drug use inclinations or habits. Thus, the significant relationships between social
network variables and substance use observed in the cross-sectional studies may not have
been caused by social network influencing substance use. Brown, Vik, and Creamer
(1989), for example, suggested that adolescent drug relapse can be caused by "situational
self-selection" (p. 299), whereby teens who desire to use drugs seek out their
pretreatment network of drug-using friends. The possibility of substance use leading
changes in the social network composition is tested in an exploratory fashion in the
proposed model (i.e. alcohol or other drug use at Time 2 is used to predict the three
network variables at Time 3). Finally, the stability of the measures is assessed by the

path from each construct at Time 2 leading to that same construct at Time 3.

39

.8389 .928 mu com: 2a aging—meson... 38252 n .82 3 d:->6=£ 5.2-8338 m.
nee; 23 9732.8 5.22.55 & «08:. .8: 85333 agave.“ :69: ESE BS8509»: A «...-arm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

98 30W 5 $8.30
053.9?“ K1 0.53.0?“—
5
~55: E82 1 3:82 «83.02
2:55... 82 33.55.... .oz
.aépasgw
85.5 ...
\ ‘i‘ 2...:
\ Ill-I- .
non: V ~33
MEG .350 9:9 .356
..e _2_8_< .5 .2302

 

 

 

 

 

 

scam—:80
.2581me

,/ ....

~32 00:33.6
83502 N€95.92 f boa—=02 453.0:
502.32 uOA Eo— aoZ ..o

e m €33. f m 3.38

 

 

 

 

 

 

 

 

 

 

2.88:0 336605 moEF £96605 N083. owcmsomaéi

CHAPTER 2
METHOD

The current study is part of the Substance Abuse and Mental Illness (SAMI)
Project, a larger National Institute of Mental Health-funded evaluation of an innovative
dual diagnosis treatment program at a public psychiatric hospital. The larger study
employed a longitudinal design with five follow-up periods. The present study utilizes
data obtained at baseline and during the first two follow-up periods.
Emisimms

The participants were consecutive admissions of men and women to Northville
Regional Psychiatric Hospital (NRPH) in Northville, Michigan. All incoming patients
were screened for eligibility based on diagnosis, problem behavior, and history of
substance abuse. Eligibility criteria required that patients screen positively for substance
abuse problems, based on a substance use and alcohol screening instrument similar to the
Michigan Alcoholism Screening Test (Selzer, 1971). In addition, patients must not have
been acutely psychotic and must verbalize some motivation to receive treatment for their
substance abuse problems. Patients were excluded from consideration if they were
severely cognitively impaired, were developmentally disabled, or if they had shown a
predisposition toward violence or inappropriate sexual behavior, which might cause

problems on one of the experimental wards that house both men and women on the same

40

41

ward. Participant recruitment occurred from June 1991 to June 1992 and the last follow-
up interview was conducted in April 1994.

Out of 2,806 incoming hospital patients during the study period, 531 patients met
the initial eligibility criteria and agreed to research participation. Further eligibility
criteria for the study were developed and required that patients spend at least five days in
treatment, which allowed all baseline measurements to be performed and assured a
minimum level of treatment. This additional criterion was applied shortly afier
recruitment began when it was observed that some participants were being discharged
abruptly after receiving a court hearing to determine if their involuntary commitment was
justified. Almost all of the patients at NRPH were involuntarily committed based on
meeting one of the following criteria: (1) being a threat to themselves, (2) being a threat
to others, or (3) being unable to care for their basic needs. Additional participant loss
occurred if respondents dropped out of the study before completing baseline
measurements. Thus, there were 467 individuals with complete baseline information. Of
that number, 395 individuals were successfully interviewed for the 2-month follow-up
(84.6 percent) and 391 individuals at the 6-month follow-up (83.7 percent).

Detailed information on the demographic characteristics of these 467 patients is
featured in Table 3. Participants were predominantly male, African American, and had
graduated from high school. Based on NRPH discharge records, the most common
diagnoses were: 27.6 percent schizophrenia and related, 8.4 percent major depression, 6.0
percent bipolar depression, 15.0 percent mild affective disorders, 21.0 percent organic
mood disorders, 95 percent antisocial personality disorders, and 16.1 percent adjustment

disorders. Determination of these psychiatric diagnoses was made by the psychiatrists at

42

NRPH using their clinical judgement. The hospital does not use a standardized
diagnostic instrument, nor did the research team independently validate these diagnoses.
According to participant self-report, 32.5% had no prior psychiatric
hospitalizations, 30.1% one or two hospitalizations, and 37.3% had three or more.
Regarding substance abuse treatment, 33.6% reported never receiving any prior
treatment, 37.9% had received either residential alcohol or drug treatment, and 28.5%
reported receiving both alcohol and drug treatments. The substances that caused the
primary problems leading to admission were as follows: Alcohol and another drug (32.0
percent), alcohol alone (31.6 percent), cocaine (21.1 percent), two or more drugs (11.8
percent), heroin (1.5 percent), and 1.2 percent other drugs (e. g. cannabis or sedatives).

Finally, less than one percent said that they did not have a drug problem.

 

 

 

Table 3. I,
mographic Variable Value
Mean age in years x Standard Deviation (SD) 33.4 :1: 7.3
Sex
Male 74.3%
Female 25.7%
Race/Ethnicity
African-American 76.2%
Non-Hispanic White 23.3%
Hispanic .2%
Native American 2%
Median Years ol'education at SD 11.20 t 2.30

Employment pattern

Full-time 30.3%
Part-time 22.5%
Retired/disabled l 0.2%
Unemployed 33.2%
Other 3.7%

Marital status

Never married 62.9%
Married 8.9%
Separated. divorced 26.1%
Widowed 2.1%

Percent ever charged with any crime
Yes 73.2%
No 26.8%
Experimental Condition

Referred to experimental Wards 64.7%

Referred to usual care Wards 35.3%

 

n = 467: Sample sizes for variables range from 311 ( due to missing data on race) to 467.

44
Design

An overview of the design of the current study is given in Table 4. F ollow-up
interviews were conducted at two and six months post-discharge. The 2-month follow-up
period was chosen because the first follow-up needed to occur shortly after discharge for
detecting short-term treatment effects. In addition, it would take approximately two
months for participants' personal networks to stabilize after their release from the
treatment facility. Furthermore, having a short time between discharge and the first
follow-up period helped the project maintain contact with participants, thereby improving
the retention rate. The 6-month follow-up period was chosen because the majority of
relapse episodes occur within the first six months following treatment completion (Hunt,
Barnett, & Branch, 1971 ).

Although not a focus of the present study, the larger investigation employed a
randomized experimental design to evaluate the impact of the specialized dual diagnosis
treatment program. Afier recruitment into the study, project staff administered a baseline
interview and then randomly assigned patients to either the specialized Mental Illness -
Chemical Dependency unit (MI-CD) or to a usual care unit. The baseline interview
contained the Addiction Severity Index (McLellan, Luborsky, Woody, & O'Brien, 1980).
Patients were given the second part of the baseline interview, which contained the social
network questions and all other baseline measures, within three days. The decision to
split up the interview was based on pilot test results that indicated the duration of the full
interview was too long for most participants. The final in-hospital interview was

completed before the participant's discharge, or at the end of four weeks afier initial

45

admission, whichever came first. Data collected with the discharge interview were not

used in the current study.

Table 4

E'EIEIE' 'Sl

 

Psychiatric Hospital Portion (Time 1)

Experimental Baseline

Condition Interview

(Part II)
I
Baseline
Random
Interview
Assignment
(Part I)
‘A

Usual Care Baseline
Condition Interview

(Part II)

Referred to
experimental

ward

Referred to usual

care ward

Discharge

Interview

Discharge

Interview

Community Follow-up

(Time since discharge)

Time 2 Time 3
(2 months) (6 months)
Time 2 Time 3
(2 months) (6 months)

The project guaranteed participants that all interview and follow-up location

information would be kept strictly confidential and that they would be paid for their

participation. Individuals were compensated: $10.00 when they had completed the

second part of the baseline interview, $15.00 afler completing the discharge interview,

and $20.00 after completing each of the five follow-up interviews. All of the study

procedures were approved by human subjects review committees at Michigan State

46

University and at the Michigan Department of Mental Health (See APPENDIX A for the
consent form). Those who indicated that they would be willing to participate were asked
to sign an informed consent statement and were then given the first portion of the
baseline interview. After the baseline interview was completed, project staff gave each
patient an envelope that contained a random assignment card to either the experimental
condition (MI-CD) or to a regular NRPH acute admission ward with a ratio of 3:2
experimental participants to usual care participants. As mentioned earlier, the
experimental design is not a critical component of the current study; however, the two
conditions are briefly described for descriptive purposes.
IheExnerimemaLILeatmenLEmgtam

The MI-CD unit provided specialized, co-educational treatment on two 30-bed
wards. Similar to other reports (Minkoff, 1989), the program philosophy endorsed the
belief that both mental illness and chemical dependency are bio-psychosocial illnesses.
The program incorporated comprehensive treatment modalities addressing physical,
psychiatric/psychological, family, social/interpersonal and substance abuse problems.
The Ml-CD program was didactic and intensive--structuring 16 hours of the patient's
weekday with scheduled therapeutic activities, such as psychiatric treatment,
individual/group psychotherapy, family therapy (as needed), 40 educational lectures,
group discussion and self-help groups on the ward, family education, activity therapy,
medical services and afiercare planning.
IhellsuaLCauLEmgmm

Usual care participants were assigned to one of the acute care admissions wards at

NRPH depending upon availability. Treatment on these wards featured activity therapy,

47
individual and group psychotherapy, and access to hospital-wide AA and NA groups.
Usual care participants received the same treatment provided to the other non-substance
abusing patients at NRPH. This treatment approach was identical to the way in which all
dual diagnosis patients were treated before the creation of the MI-CD unit. Treatment
staff on the usual care wards did not receive any special training in substance abuse, in
contrast to the specialized training received by MI-CD unit staff. Finally, the usual care
wards had a lower staff-to-patient ratio than the MI-CD unit.
Communitxiollnmunllmdum

Follow-up interviews were conducted at 2-, and 6-months post-discharge. As
mentioned earlier, most relapse episodes occur in the first six months of treatment
completion, making this an ideal period for studying the social factors influencing the
relapse process. Experimental participants spent an average of 50.8 (SD = 32.1) days in
treatment, compared to 30.8 (SD = 25.4) days for usual care participants.

Given that previous studies in the dual diagnosis and substance abuse literature
have been marked by high attrition, extensive efforts were employed to ensure high rates
of follow-up interview completion. Procedures for locating participants in the
community were similar to approaches employed by other researchers conducting
longitudinal studies (Cohen, et al., 1993; Davidson, Redner, Amdur, & Mitchel, 1990;
Goldstein, Abbott, Paige, Sobel, & Soto, 1977; Rumptz, Sullivan, Davidson, & Basta,
1991). The project's general tracking approaches were to: Make contacting the project
convenient and enjoyable for each participant, reward participants when they contacted
the project office by phone or mail, contact participants frequently, reward interviewers

for locating participants and completing interviews in a timely fashion, and dedicate

48

substantial resources to tracking. The population in this study was generally considered
difficult to locate because some participants had sparse tracking information due to
having few friends and relatives or because they did not have a stable residence.

Tracking information and release of information forms were collected at baseline
and confirmed or updated at discharge as needed. Two release forms were completed,
one for friends and relatives and another for formal service agencies. For the first release
form, respondents provided the names of friends and relatives that project staff could
contact in the event that the interviewer had difficulty finding them. The other release
form was developed in conjunction with the Detroit-Wayne County Community Mental
Health Board (D-WCCMHB) to allow access to service utilization records for the
purpose of locating any difficult-to-find participants. If interviewers experienced
difficulties in locating a particular participant, the D-WCCMHB representative would
identify the most recent treatment services utilized and then contact the agency or
agencies that provided the services in order to obtain more information about the client's
location. Interviewers also probed to generate the names of places where the participant
may have had criminal justice system involvement, where they have received previous
medical and psychiatric services, the churches they attended, and the shelters frequented.
New release and tracking forms were completed after each of the two community follow-
up interviews.

If the relatives/friends were reluctant to help locate the participant and the
interview was at least two weeks late, they were offered $10.00 to $25.00 for their help in
coordinating the interview. In addition, interviewers were also offered monetary

incentives for completing interviews in a timely fashion. If the participant could not be

49
found by the previous approaches, more formal procedures were employed, such as
examining the hospital's discharge records and checking the respondent's service
utilization patterns. Finally, as a last resort, a list of "lost" participants was distributed to
the local morgue on a regular basis to help locate deceased participants.
Measures

An overview of the measures and data sources is provided in Table 5.
Psychometric information (i.e. internal consistency) based on the present sample is
provided for most study measures. In addition to assessing the internal consistency of
scales in this study, the inter-rater agreement for these scales was also computed for 45
interview pairs in which a project staff member coded the interview while the primary
interviewer questioned the respondent. Because more sophisticated analyses are
ongoing, the agreement statistic presented here is based on the percent agreement
between pairs; percent agreement was not corrected for chance agreement. However, the
typical rates of agreement are sufficiently high (averaging 2 98%) that even chance-
corrected measures would certainly be acceptable.

AddimiQmSexemJndex (McLellan, Luborsky, Woody, & O'Brien, 1980) The
ASI is a structured clinical and research instrument that assesses the severity of seven
unique problem areas commonly found in patients with alcohol and substance problems.
These seven domains are: Medical, Employment, Legal, Drug/Alcohol, Family/Social,
and Psychiatric. However, the medical, employment, legal, and family/social sections
were not used in the current study. For the first section of each problem area, participants
are asked several objective and potentially verifiable questions about their problem

symptoms. In the latter portion of the section, participants provide subjective

50

information about the extent to which they have been bothered by these problems and the
importance of treatment based a 5-poin't response scale ranging from 0 "Not at all" to 4

"Extremely. "

Table 5

11 lI'E'lfilElE' '51

51

 

 

 

 

Time Frame
Post-Discharge F ollow-up
Time #2 (2 months)
Baseline Measures (Time 1) Time #3 (6 months)
5 . l S E | . . [2'

Family Involvement in
Treatment (Involved,
Moderately involved, Not
involved)

Demographics.

(Possible Co-variates)

Sex

Age

Ethnicity (White/Non-white)
Years of Education

Marital Status

Employment status (typical
employment pattern over past 3
years)

E . 1:1"

Experimental (MI-CD) wards
Usual Care wards

TreatmenLQuration (Number of
days)

(Both scales standardized and combined

into one score)

- Symptom Checklist - 10 (SCL-10)

- Addiction Severity Index (ASI -
Psychiatric Composite Score

SociaLSunport
- Sobriety Support of Network Members
(Mean reported support for staying

clean)

SociaLNetmrk

. Number of intimate network members
(Total number of people mentioned in
response to first probe question) -
confidants

. Network Substance Use (Proportion of
network members reported to use drugs
or alcohol)

AlthLandmthDmgflse

. Alcohol Relapse (Use of alcohol to
intoxication in the 30 days prior to the
interview)

. Drug Relapse (Any use of 12
substances in the 30 days prior to the
interview)

- ASI Alcohol Composite score

. ASI Drug Composite score

 

52

The ASI yields both W385 indicating need for additional

treatment and also W that are indicators of overall problem severity and
are capable of showing change over the last 30 day period. The Interviewer Severity
Ratings were designed to be used for treatment planning and the composite scores for
research purposes. Therefore, the ASI composite scores were used in the present study.
In this study, for individuals who had mostly complete data (66 to 75 percent
complete), median substitutions replaced missing scores. Composite scores were
composed of items that referred to the prior month, which allows these scores to measure
change. For example, the alcohol composite score was derived from six questions based
on the prior 30 days: (1) numbers of days alcohol use, (2) number of days alcohol used
to get drunk, (3) number of days experiencing alcohol problems, (4) the amount of
money spent on alcohol, (5) how troubled the person was by alcohol problems (ranging
from 1=Not at all to 5=Extremely), and (6) how important treatment is for their alcohol
problems (same 5-point scale). The responses to each item are divided by the product of
the highest possible score for that item and the total number of items composing that
particular composite score. For instance, the number of days of alcohol problems is
divided by 180 (30 days X 6 items). It should be noted that the amount of money spent
on alcohol can have extreme values with non-normal distributions; therefore, the scale
developers adjust for these responses by using a log normal value for the response. The
resulting composite scores range from 0 to 1.0 with higher scores indicating more serious
problems. Each of the composite scores; however, is scaled differently and therefore

scores are not directly comparable across the different domains. For example, a score of

53

.14 on the Drug Composite does not necessarily equate with a score of . 14 on the Alcohol
Composite.

In the present study, the percent agreement between matched interviewer pairs
was 97 percent for the alcohol and drug sections of the ASI and 96 percent for the
psychiatric section. The original report documents that the ASI Alcohol and Drug
Composite scores were significantly correlated with the number of overdoses, total years
of regular use of alcohol/drugs, and the amount spent on alcohol/drugs per week. The
average concordance between trained interviewers is .89 and the interview has adequate
test-retest reliability. Stronger evidence for the concurrent validity of the ASI has also
been provided (Kosten, Rounsaville, & Kleber, 1983). However, some caution is
warranted when applying this measure to persons with dual diagnoses because this
measure has not yet been validated with this population (Ridgely, Osher, & Talbott,
1987)

Sympmmfiheckflw (Nguyen, Atikisson, & Stegner, 1984). The SCL-IO is a
10-item version of the SCL-90 (Derogatis, Lipman, & Covi, 1973) that is used as a global
index of psychopathology or discomfort. The SCL-10 was originally developed based on
the findings of factor analysis of SCL-90 scores which suggested a five factor solution;
however, three factors were the most clearly defined and accounted for the most variance.
Therefore the items constituting the SCL- 1 0 represent three factors (Depression,
Somatization, and Phobic Anxiety). The reliability of this scale in the current study was
.91 (coefficient alpha) and in study of 3,628 individuals receiving service treatment was

.88. There was 96 percent inter-rater agreement for this measure.

54

W. The approach to measuring the

participant's social networks was based on the approach of McCallister and Fischer
(1983). This methodology, which uses probe questions to enumerate the network, is
commonly used with general population samples, but has also been recommended for
research with schizophrenics and other special populations (Phillips, 1981).

Five probe questions were used to elicit core network members. The first
question was "When you are concerned about a personal matter--for example, something
you are worried about or you are concerned about someone you are close to--who do you
talk with?" This question, originally developed for use in the General Social Survey by
Burt (1984), has also been used with chronic mental patients (Holmes-Eber & Riger,
1990). One difficulty reported by the latter authors was that several individuals were
unable to think of anybody and therefore two additional probes needed to be developed.
The four other questions used in the current study asked participants who they spend time
with, who they would borrow money from, if they have good friends in either an AA or
NA group, and finally if there is anyone else important in their life that they have not
mentioned. After the network has been enumerated, participants are then asked to list the
first name and the first letter of the last name of each network member, a procedure
which still preserves the confidentiality of the alters.

After the list of names is compiled, the name interpreter questions are asked. For
example, detailed background information is asked about each person including
demographic questions (i.e. sex, relationship, and whether they met the individual at the
psychiatric hospital), the frequency of contact with the network member, how important

this person is to the respondent, and how satisfied the participant is with their contacts.

55
None of the previous name interpreter questions were used in the present study. Of
greater interest is the proportion of the total network that drinks alcohol or does drugs
(Network Member Substance Use), which is calculated by dividing the reported
substance use of the members by the total number-of network members elicited from the
five probe questions. Furthermore, the support the individual receives from network
members for helping the person to stay clean and sober (Sobriety Support) is simply the
average rating across all reported network members. The inter-rater agreement was 97.8
percent in the present study. See APPENDIX B for a copy of the network measure used
in this study.

Wm The extent of family involvement in treatment
was coded based on hospital records. Project staff referred to pre-release discharge
summaries and psychosocial summaries recorded by social work staff and coded the
extent of family involvement into one of three categories: Involved in treatment (60.3
percent), only moderately involved (18.7 percent), and not involved (20.5 percent). The
psychometric properties of this rating scheme were not assessed.

Demographics, Records kept by the psychiatric hospital were accessed for basic
demographic information (e. g. ethnicity, educational level, date of birth).

Data Analysis

First, a series of analyses were conducted to compare those individuals with
missing data on one or more of the 25 study variables (approximately 180 cases) with
those individuals having complete longitudinal data for Baseline, Time 2, and Time 3

data. These analyses help provide an understanding of potential biases resulting from

56

nonresponse error and are commonly reported (Wasylenki, Goering, Lemire, Lindsey, &
Lancee, 1993, for example).

The panel model proposed in the introduction will be tested using path analysis.
The path coefficients for this model were determined using hierarchical multiple
regression, which allows variables to be entered in blocks specified by the user. Any
demographic variables significantly correlated with the dependent variables were used as
covariates that were entered in the first block along with experimental group condition,
length of stay, and family involvement in treatment. In the second block, for example,
variables were entered in the following order: Mean level of Sobriety Support, Network
Member Drug Use, Intimate Social Network Members, and Psychiatric Distress.

There are four: different dependent variables and therefore four different sets of
regression analyses. First, separate analyses were conducted for alcohol and for other
drugs, which is fairly common in the substance abuse literature (see Ellickson & Hays,
1992). Furthermore, for both substances, analyses were conducted for relapse (based on
the 30 days prior to the 6-month follow-up interview) and also for the extent of substance
abuse problems as measured by the Addiction Severity Index. Logistic regression was
used for predicting whether relapse occurred, whereas linear multiple regression was

used for the ASI composite scores.

CHAPTER 3
RESULTS

Several analyses were conducted to test the hypotheses about the relationships
between substance use outcomes and social networks, social support, and psychiatric
distress. A brief overview of the data analysis plan is provided here. First, descriptive
statistics are presented on the social network variables at both follow-up periods. Next, a
series of analyses were conducted to compare individuals with missing data with those
individuals having complete data on all study variables. A correlation matrix of all 25
study variables is then presented and discussed. The hierarchical multiple regression and
logistic regression analyses that were used to obtain the standardized path coefficients for
the model are then presented to provide more detail about the relationships between the
predictor variables and the substance use outcome variables. Figures of the four path
models, trimmed of nonsignificant pathways, are then presented. Finally, the significant
direct and indirect pathways leading to substance use at follow-up are discussed.

Descriptive characteristics of the social networks at follow-up are provided in
Table 6. The social network variables were fairly stable from the 2- to the 6-month
follow-up. The typical social network was fairly small in size. The social networks of

this dually diagnosed sample were composed of approximately five members at both the

57

58

2- and 6-month follow-up. Eleven participants (2.8 percent) at both follow-up periods
were complete isolates, that is they reported having no members in their social network.
In contrast, a subset of over 10 percent of the participants reported having networks
composed of 10 or more members at the first follow-up .

The number of intimate network members was based on responses to the first
network enumeration question, which asked about the number of individuals with whom
the respondent can discuss their personal problems. About two of the five network
members at each follow-up period were intimate members, or confidants. Relatives
constituted just over half of the network members, on average. At both time periods,
over 40 percent of respondents had networks composed of at least two-thirds relatives.

The majority of network members were not users of alcohol or drugs. On
average, about one-fourth of the network members were reported by the respondent to
consume alcohol or drugs. Thus, the typical network would have about one alcohol or
drug user out of the five network members. A high proportion of networks were
composed entirely of "clean and sober" members at both the 2-month (44.8%) and 6-
month (39.6%) follow-ups. F urthermore, less than six percent of the networks were

composed entirely of alcohol and drug-using members.

59

 

m.m _N .m w... 2 .. m
..n om .. we hm v
_.m_ on m ad. 5 m
_.mm «.2 N new c2 N
Won 2. _ ONM hm. _
o»... a ma..— wd. 9. o w?— “ Na... ”N. _m o .3353». 2.22:5... .62
are ..m to. _.o_ o... 5.
9». I o ad __ o
..m cm x oh cm M.
Na mm N. m... om h
We. S. c x... mm c
...: S. m ..2 we m
...: we .. 9N. om ..
...: mm. m 0.3 3 m
We. 2. N ..Z I. m
...m _N _ 0.5 on _
E: a mm... .3 __ o Ia .« was... w.~ __ o "Sm 4.02.32 5:33
Om H :32 3. x2532... 92.5 Om a .892 ...... Sconce...— o.._..> 933:5

 

m 2......

 

N 25..

 

 

o 93.;

60

.8030... .8082. 3:8... :00 0... 80...: 5.3 0.000.. - 3.50:0 5.8.0880 ...03.0: .8... 0... :0 003.. .

 

 

 

 

on 0. 8.. 0.... ... 8..
.... 3 00...... m0 mm 00.4....
...... S 00.4%. ...m. .2. 8.4%.
....on m: 2.-.... ....em .3. 3.-....
0~.o a cm... 0.0... cm. so... mad H «No «.3. a... cod 8003 man. .0 .0..00_< 5.:000...
...... cm 2... 0.... no 8..
..0m .... 00...... ...cm 8. 00.4.0.
0...~ co. coin. Nam 00. 00.4.0
«4... on 3.0. ...c. S mm.-.o.
.....o H and ...N. .... cod .2. H 000 .... m... cod m0......3. ..0.:0:0.m
Om « 8.02 a... 55:00.... 0:.:> Em A 8.03. .... ...—.0200... 0:.:> 0.8....5
m 08.... N 08....

 

250. 0 2...:

61

l l' . E i I

In this study, missing data was caused by participant refiisals, participant deaths,
and the inability of the interviewer to locate the participant. Most of the attrition was
caused by the latter cause, which is typical of most panel studies dealing with similar
populations (i.e. persons with substance abuse or psychiatric problems). Eight
participants were deceased over the course of follow-up data collection based on reports
of relatives and records maintained by the Wayne County Morgue and the Michigan
Department of Public Health. Substantial amounts of missing data can reduce sample
size, thereby lowering statistical power, and also introduce nonresponse error if
participants with missing data differ from participants with complete data. Therefore, a
series of analyses were undertaken to provide a better understanding of potential sources
of nonresponse error in this investigation.

0f the original 467 participants interviewed at baseline, 395 and 39] were
interviewed at Time 2 and Time 3, respectively. In addition to unit nonresponse caused
by this participant attrition, item nonresponse was caused by respondents refusing to
answer particular interview questions. Moreover, item nonresponse can be caused by
interviewers neglecting to record participants' responses and also by errors made during
data keypunching. One significant source of item nonresponse in this study was
incomplete data (53 cases missing) for the race variable that was coded from existing
NRPH records. Finally, some participants were interviewed only at the 2-month follow-
up and not the 6-month follow-up, or the reverse, which further contributed to listwise

missing data.

62

 

 

 

 

 

Table 7
OOIOoIOtéllOI°.cIOaO t, 0H0"-0'U. “Lou'wn'ut
andfiuhstancellselariahlfi
Level of Data Completeness'
Missing 1 or Complete
more variables Data
11 = 184 n = 283 Totals”
(394%) (60.6%) (N = 467)
n % n % n %
Sex
Male 137 74.5 210 74.2 347 74.3
Female 47 25.5 73 25.8 120 25.7
x3 (l) = .00, n.s.
Race
White 31 23.7 66 23.3 97 23.4
African American 100 76.3 217 76.7 317 76.6
x3 (1) = .01, n.s.
Marital status
Not married 165 90.7 259 91.5 424 91.2
Married 17 9.3 24 8.5 41 8.8
x2 (l) = .10. n.s.
Emplopnent status
Not employed 89 49.4 126 44.5 215 46.4
Employed 91 50.6 157 55.5 248 53.6
12(1)= 1.07, n.s.
Experimental condition
Usual care 71 38.6 94 33.2 165 35.3
Experimental 1 13 61.4 189 66.8 302 64.7

x:(l)= l.4l,n.s.

63

 

Level of Data Completenessa

 

 

 

Missing 1 or Complete
more variables Data
11 = 184 n = 283 Totals”
(39.4%) (60.6%) (N = 467)
n % n % n %
Age
Mean 1 Standard Deviation (SD) 33.8 i 7.5 33.1 :t 7.1 33.4 t 7.3
E (l, 465) = .90, n.s.
Treatment duration
(Number of Days)
MeaniSD 44.8i388 43.2i25.8 43.8i3l.4
E (l, 456) = .27. n.s.
ASI Alcohol Composite at baseline
Mean :1: SD 0.41 i 0.25 0.41 :1: 0.26 0.41 10.26
E (1, 464) = .00, n.s.
ASI Drug Composite at baseline
MeaniSD 0.17:0.11 0.201013 0.19:0.12

E (l, 464) = 6.22, n <.05

 

"There were 25 study variables: individuals with valid values for all 25 variables were considered
to have complete data. "Totals for some variables may be less than 467 as a result of missing

data.

64

Participants were divided into two groups depending upon the completeness of
their interview responses. There were 283 individuals who had complete data for all of
the 25 study variables; 184 respondents had missing data for one or more variables.
Comparisons between these two groups were made using Chi-square analysis and
oneway ANOVAs. These analyses, featured in Table 7, revealed no significant
differences between the two groups on the five demographic variables (i.e. sex, race,
marital status, employment status, and age). Furthermore, experimental condition and
treatment duration were unrelated to level of data completeness. Thus, individuals in the
usual care condition or who spent more time in treatment, possible indicators of poorer
prognosis, were no more likely to have missing data.

Two substantive variables, baseline alcohol and drug problems, were also related
to the missing data variable. Surprisingly, individuals with greater drug abuse problems,
as measured by the ASI, were more likely to have complete data (F(1, 464) = 6.22, p <
.05). However, the magnitude of the difi‘erence in ASI scores was quite modest (.17 i
.11 versus .20 i .13).
lntnmmlatiQnLAanthLSLudmmhles

The zero-order correlations among all 25 study variables are featured in Table 8.
The intercorrelations among the demographic variables are displayed in the first triangle
(Variables l - 6) off the diagonal. The next triangle is composed of the predictor
variables (7 - 17) followed by the substance use outcomes (18 - 25). The sample size for
this matrix was n = 283 because correlations were calculated using listwise deletion of
cases. Therefore, correlations were statistically significant with absolute values greater

than r > .098 (.05 level, one-tailed) and r > .14 (.01 level, one-tailed).

65

As part of the analysis plan, any demographic variables that were significantly
correlated to the substance use outcomes were included as covariates in the regression
analyses. Those significant correlations are discussed here. Being married was
associated with lower ASI alcohol composite scores at both follow-up periods. The ASI
alcohol composite at Time 3 was related to both gender and age, with males and older
participants having greater alcohol problems. Married individuals were less likely to
relapse using alcohol. Educational level was associated with a greater likelihood of
alcohol relapse at Time 2 and marital status was the only variable related to Time 3
alcohol relapse.

Women had greater ASI drug composite scores at both follow-up periods and
African Americans had greater ASI drug scores only at Time 2. Furthermore, at Time 2,
ASI Drug composite scores were lower and the probability of drug relapse lower among
individuals whose typical pre-admission employment pattern was to be working part- or
full-time. Married participants were less likely to relapse with drugs. At Time 3, the
only demographic variable related to either ASI Drug composite scores or drug relapse

was age, with older participants more likely to relapse.

66

602337. u§<uo 02.33. 9.5 .28 .0400? ”1.5.5.003“. 0.00 15.5.1.0 00.0.2.0... E253... doth-Eu. 005-... 2.5.3... 82".. .552 .vo.A0...Emn. 2.3.0.3 82.1.0

 

2058.30 28.55 38¢... 323.8 8.: v....)... «15.8 3 a. 33...: 36.5.9.0... 82.. ...... 0.88.0 8.68... 2F .05. 2.2 8.3.: 8 v0 2.. a .598»... a. z A .. ...: :25... .26. 8. v... 2.. .. .5889. a. 08 A .. .8" am .37.

 

 

 

 

 

 

 

 

.. 0v. 8. 8. 3. 8. 2. 8. 8. 8. 8. 8.. 8. 8. 8.. 8.. 2.. 8. 8. 8.- 8.. 8. 2. 8. 8.. 31850098 .2. 8
.. 8. 2. 8. 3. 8. 2. 8. 8.- 2. 8.. 2. 8. 8.. 8. .... 8. 8.- 8. 8.. 8. :. 8. 8.. 88.3.8.0 .8
-. 8. 2. 8. 8. .4. 8. 8. 8 8.. a 8.. .n 2.. 8. 2.. 2.. o... 8 8. ... 8. :. 3.8950 .282 .9. 8
.. 8. 2. .n. 8. 2. 8. 2. 8.. 8. 8. 8. .. . 8.. 2.. 8- .... 8 8. 8.- 8. 8. 3838.282 8
.. 8. 9. 2. .n. 8.. 8. 8.. 2.. 8.. 8.. 8 2.. 8. 8.. o... 2.- 8 8.- 8 8. 3.898002: .9. a
.. 2. 2. 2. 8.. 2. 8.- 8. 2.. 8. o... 8.. 8 2.- 2.. 8.- 8.. 8, 8. 8. ugomg 8
.. x. 2. 8.. 2. 8.. ... 2.. 2. 8.. 8.. 8.. 2.. 2.. 8. 2. 8. 8. 8. 8.880.282.0< 2
.. 2. 8.. 2. 2.. z. 2.. 8. 2.. 8.. 2.. o... 8.. 8.. 2. 8. 8.. 8, 88.50582... 2
-- 8.. 8. 8.- 8. 8.- 8. 8.. :.. 8. 8.- 2.- 8.. 8.. 8 8. 8.- 2.23: 9.5.2.2... 2
.. 8. 2. 8. R. 8. 8 8 8. 8. 8.. 8 8.- o. . 8.- o. . 33.8: 28...... oz .2
.- 8 . 8. 8 . 8. 8 - 8 8 . 8. 8 8.. 8 - 8 8 . 8 8.: 3...»: 82.2 .2
-- 8.- 2. 2.. 8 8. 8.. 8.- 8. o. o. 2.- 8. : . 9.888 0.58 I
.. o..- 8.. 8.. 8.. 8. 8.- z . 8.. 8 8 8 8.. ~33: 9.5.3.05 2
.- 8, a 8.. 8. 8 8.- 8. 8 :.. o... 8. 33.52 225.... .oz .2
-- 2 . 8. :.. 8. 8 8. 8.. 8.. 2.. 8 Na: .35.: 82.2 :
.. 8. 8. 8. 8. 8 8 8.. o. 8- 85.2.8 22.28 8.
-- on . 2 - 8 8. o... :.. 2 . ... - .6523... 8.5. ....
-- z. 8. 8- 8 :. 8. 8 8.5.35.3. oz 0
-- 8. .8 8. 8.. 8. 8.. .5380 .C s
-- 8.. 8 - 8. 8.. 8 - . 3.52 8
-- 2 8.- : - 2 2558.850 8
-- 2 8 8.. 888...... .4
-- :. 8. on... m
.- 8- .8... N
z .50..
8 8 8 8 .N 8 2 ... ... 2 2 z 2 2 : 2 o 0 .. e n . m N .
1.10M 4.. .W 2; . .0 .. . .1. .-...> : .4 4. ...-.. , . D.. o . .454 . m0 .QtflN’ u .2 : 1. -m .0 m eta—0:0 H . 1.415 .4 .m 0—DQH.

67

Among the predictor variables, individuals randomly assigned to the experimental
ward spent more days in treatment and experienced fewer alcohol problems at follow-up.
Many of the social support variables were significantly correlated, indicating some
overlap among these constnrcts. For instance, individuals reporting greater amounts of
sobriety support also had fewer substance users in their networks and a larger number of
intimate network members. However the magnitude of most of these relationships was
usually rather modest (most is < .10).

Of the predictor variables, psychiatric distress ‘was most consistently related to the
substance use outcome measures. Psychiatric distress at Time 2 was significantly related
to three of the four cross-sectional substance use outcomes (the exception being alcohol
relapse at Time 3) and all four prospective outcomes. Psychiatric distress at Time 3 was
significantly related to all four cross-sectional substance use outcomes. Some of the
social support/network variables were significantly correlated to outcomes, but less
consistently and of smaller magnitudes. For instance, individuals with greater family
involvement had lower ASI Drug composite scores and individuals with greater sobriety
support were less likely to relapse with alcohol. Having substance using network
members at Time 2 was positively related to ASI Alcohol Composite scores and an
increased probability of alcohol relapse for both follow-up periods. Finally, the number
of Time 2 intimate network members was negatively related to Time 2 alcohol and drug
relapse.

The intercorrelations of the substance use measures suggest that the ASI Alcohol
and the ASI Drug Composite scores were fairly independent constructs, but there is

considerable overlap between the two indicators (ASI scores and relapse) of each

68

substance. For example at Time 2, the correlation between ASI Alcohol Composite score
and alcohol relapse (r = .54, p < .01) was higher than the correlation between alcohol
relapse and either drug relapse (r = .19, p < .01) or alcohol relapse and the ASI Drug
Composite score (1 = .16, p < .01).
WW

Standardized path coefficients for the demographic, social network, and social
support measures used to predict substance use were obtained using hierarchical multiple
regression (see Cohen & Cohen, 1983 ). All of the path coefficients are presented for the
four difi‘erent versions of the model tested. These coefficients are presented in Appendix
C (Tables C-l to C-4) because the figures (presented later) are trimmed of all
nonsignificant paths and because it is not essential to present these data in the results
section. Data from these tables are not discussed in detail because study findings relating
to the ASI composite scores and relapse are discussed in the next section describing the
regression analyses.

Path analysis was conducted using hierarchical multiple regression and logistic
regression. Two-month substance use measures (Time 2) were predicted using two
blocks of predictor variables and 6-month substance use (Time 3) was predicted with four
predictor blocks. The first block contained any demographic control variables and the
pre-discharge variables (experimental condition, treatment duration, family involvement
in treatment). Recall that it was hypothesized that more positive outcomes would be
associated with receiving the experimental condition, receiving longer treatment, and
having greater family involvement in treatment. The second block contained the social

network/support variables (sobriety support, network member substance use, number of

69

intimate members) along with psychiatric distress. It was hypothesized that network
member substance use would be the best predictor of outcomes among the social
support/network variables and that greater psychiatric distress would lead to poorer
outcomes. The third block for the 6-month substance use variables consisted of 2-month
substance use, which controls for the autocorrelation between the two variables. Finally,
the fourth block was composed of the 6-month (cross-sectional) social network/support
variables and psychiatric distress. The advantage of using hierarchical regression was the
ability to control the impact of pre-discharge variables and determine whether the social
variables and psychiatric distress provided additional predictive power at each step.

Winn. Table 9 contains the hierarchical regressions for
the ASI Alcohol and Drug Composite scores at each follow-up period. It should be noted
that the beta weights featured in this table are fi'om the full model after all variables have
been entered and not the beta weights after each step. Regression results for the Time 2
ASI Alcohol Composite scores (first column in Table 9) are discussed first.

None of the pre-discharge variables (i.e. marital status, treatment condition,
treatment duration, family involvement) in the first step emerged as a significant
predictor of the Time 2 ASI Alcohol Composite scores. Nonetheless, the R2 (.03) of the
complete predictor set was significant (p < .05). The block of social network/support
variables and psychiatric distress, which were entered in the second step, significantly
contributed to the R2 of the equation (R2 change = .19, n < .01) predicting the ASI
Alcohol Composite score. Within this block, Time 2 Network member substance use (D

= .20, p < .01) and greater psychiatric distress (B = .37, n < .01) were significantly

70

related to ASI alcohol scores after controlling for all pre-discharge variables that were
entered in the first step.

For Time 3 ASI Alcohol Composite scores, age ([3 = .10, p < .05) and treatment
duration (9 = -. 17, p < .01) emerged as significant predictors, with older participants and
individuals spending less time in treatment reporting more alcohol problems. In the
second step, network member use ([3 = .09, n < .05) was related to greater alcohol
problems. The ASI Alcohol score at Time 2, the only variable entered in the third step,
was highly related to the ASI score at Time 3 ([3 = .55, n < .01). In the fourth step,
greater psychiatric distress was associated with higher ASI alcohol scores ([5 = .26, n <
.01), even after controlling for the pre-discharge and social network variables, and Time
2 A81 Alcohol Composite scores. The final model predicted a respectable amount of

total variance in Time 3 ASI alcohol scores (R2 = .43, n < .01, Adjusted R2 = .41).

 

 

 

 

 

 

 

 

:8. ...... :2. :8. N 3.5 .. ...
8.. 8. a 8. :..... 8. « 8. .m... 8. a 8.. :8. s. « 8. «.895 .136»...
8. 8. u 8. 8.- 8. « 8. 8.. 8. « 8. :.. S. a 8.- «825.3% 62
8.- 8. a. 8.- 8.. 8. u. 8.- .8. 8. « 8. :8. 8. « 2. «a: 3.52 «.252
8.. 8. « 8. .8.. 8. a .... 8.. 8. a. .8.. 8. 8. « 8. 9.8.5 but...“
Mann-NEH . a new
8. .8. :8. .8. . new a ... ... 8..-6
.... 8. 8. .8. . ..sm 3 ... 89.83.
8. .8. :8. .8. . 3m .- ...
8.. 8. a 8.- :.. . .- 8. a 8.- 8. 8. « 8. 8.. 8. ... 8.- .8523... 2.3.“.
8.- 8.. « 2...- 8.- 2.... a 2.... :....- 3." « 8.8. 8.. SN « .N. .. gap ... 55 .02
N 8. 8. « 8. .o. 8. a. 8. 8.. 8. a 8.8 8.. 8. « 8.- .8....zoo .8335
.. .. .. .. 8.. 8. « 8.- 8.. 8. « 8.. .....52
.. .. 8.. S. a 8.. .. r r r .3353
.. .. .. u .. .. .. .. 83m
.. n r .. 8.. 8. a 8. .. .. on...
.. .. .. .. .. .. r .. .8....
.. .. .. r 8. 8. « 8. u .. .5
Egg . . mam
a 83 8 mm... a uni. ._ mm: 88...»...
azusnna... 38.535 .88.... nos: 88...: «8.:
3.89.8 9.5 .9. 3.8980 .382 .8.
.... ...... ...}...A .. .... ....a...:.1. .. .... .. .35.... . . . ... ...... a . ... .0... : ...... ...... do—nah.

72

3. v do. no. v do

3.03.... ...-15.889385} 3.133%». 26 15888.18 3312.8 .352...

3.58 2.85.. 32.... 3.52 8.8313433 82".. .8333... ...-.55.. 339?. 323.88.... 63...... ...-3883 N.. 8..-33832.... 85.3388... 8.32... B...
gage-853838.82 ggiéiufiwfiaiggfigigi 32.133333833838323. 9%.

 

 

 

 

 

 

...... .. ..8. .. ... ... 2.56 .3...
8. .... .. ... 35...... .3...
:..... a 3m .8 :8. n as... 8.. ... .8....
8. « 8. 8. « .... 8. a 8. 8. « 8. ...-:80
:8. 8. a 8. r .. :8. .... a 8. .. 3.5.5 91329....
8. 8. a 8. .. .. 8. .... « .... .. .. 335...... .oz
8. 8. « 8. 8.- 8. « 8.. .. .. as: 3.82 «53.02
8. 8. « 8. r .. 8.. .... « 8.. .. in 8.38
...—fig - .. 88m
... .. :8. .. n can .- ... ... 85.6
.... 8. .. 8883...???
:8. .. :..n. .. n as... .. ...
:8. 8. a 8. r a .. «.980 men .9.
.. :8. 8. « 8. .. ~95 .38.... .8.
«g . n mam
..8. ...... 8 :8. a 82m .. ... ... 85.6
8. 2. z. 8. N as... ... ... 33...!
a mmfi. .. mm... a uni. ._ mm .. 88.3...
6.7.38.8 38.5838 88.5 .8.: Ghana «8.:

 

3.83885 .9.

 

319983.?

 

73

In the third and fourth columns of Table 9, study variables were regressed on the
ASI Drug Composite scores. For Time 2 ASI Drug Composite scores, Family
involvement ([3 = -. 14, p < .01) and psychiatric distress (D = .35, p < .01) emerged as
significant predictors, with greater family involvement and less psychiatric distress being
associated with fewer drug problems. Surprisingly, none of the social network/support
variables were related to either Time 2 or Time 3 ASI Drug Composite scores, although
the increment in R2 attributed to the block was significant (R2 Change = .06, p < .05).
The only significant predictors of Time 3 ASI Drug Composite scores were Time 2 ASI
scores ([3 = .37, p < .01), and Time 3 (cross sectional) psychiatric distress (D = .36, p <
.01). The final model for Time 3 ASI Drug composite scores predicted less total variance
(R2 = .30, p < .01, Adjusted R2 = .27) than the model for the ASI Alcohol composite
scores (R2 = .43, p < .01, Adjusted R2 = .41).

The multiple regression analyses discussed thus far have addressed the
hypotheses about pre- and post-discharge variables predicting substance use outcomes.
However, a research question also was proposed whereby 2-month substance use
measures were used to predict 6-month network variables. The beta coemcients for these
paths are featured in tables C-l (ASI Alcohol composite), C-2 (ASI Drug composite), C-
3 (alcohol relapse), and C-4 (drug relapse). None of these paths are featured in any of the
four figures because none of the coeflicients reached statistical significance.
Furthermore, the magnitude of the coefficients was minuscule for the ASI variables (B's
ranging from .00 to .02) and small for the relapse variables (D's ranging from -.03 to .06).

Winks. The ability of the proposed model to predict alcohol

or drug relapse was also tested. Logistic regression, which requires fewer statistical

74

assumptions than discriminant function, was used to predict relapse. These results are
featured in Table 10. Blocks of predictor variables were entered in sequential steps,
identical to the hierarchical regression analyses. Instead of yielding statistics on the
amount of variance explained by the model (i.e. R2), the fit of the model at each step is
assessed along with its ability to predict both of the dichotomous outcomes. In this
investigation the probability of alcohol relapse was .44 at Time 2 and .45 at Time 3. The
odds of alcohol relapse are defined as the ratio of the probability of relapse to the
probability that it does not occur. Therefore, the odds of alcohol relapse at Time 2 were
.79 (0.44 / (l - 0.44)) and .83 (0.45 / (1 - 0.45)) at Time 3. The probability (and odds) of
drug relapse at Time 2 were .61 (1.56) and .61 (1.56) at Time 3. Alcohol relapse was not
defined simply as any use of alcohol in the prior 30 days (the definition used for drug
relapse), but referred to any use of alcohol to the point of intoxification in the 30 days
before the follow-up interview. At both Time 2 and Time 3, 57 percent of participants
reported any use of alcohol on one or more days in the 30 days prior to their follow-up
interview. Alcohol use to intoxication was chosen as the outcome measure instead of any
alcohol use because for participants who only needed treatment for their problem drug
use, low or moderate alcohol consumption should not be problematic.

The second column (with the heading b :l: SE) in Table 10 features the logistic
coeflicients. The logistic coefficient is interpreted as the amount of change in the log
odds of relapse that is associated with a one-unit change in the independent variable. For
example, for Time 2 alcohol relapse, the coefficient for experimental condition is -.42.
This coefficient indicates that when experimental condition changes from 0 (usual care

condition) to 1 (experimental condition) and all other independent variables are held

75

constant, the log odds of alcohol relapse occurring are decreased by .42. The third
column (R) features the partial correlation between each independent variable and
relapse.

The pre-discharge variables were entered in the first step of the Time 2 alcohol
relapse logistic regression significantly enhanced the ability of the model to predict
relapse. The model was better at predicting abstainers (84.6 percent accuracy) than
relapsers (25.8 percent accuracy). This model, using only three pre-discharge variables,
was overall able to correctly classify the relapse status of about three out of five (59.5
percent) participants. Afier the addition of the next block of variables, which contained
the social network/support and psychiatric distress variables, the overall prediction ability
improved slightly (62.7 percent), but significantly (p < .01). For Time 3 alcohol relapse,
the only significant predictors were Time 2 alcohol relapse and Time 3 psychiatric
distress. The final model for Time 3 alcohol relapse predicted 78.1 percent of abstainers,
an improvement over the base rate of 54.6 percent, and predicted 67.3 percent of alcohol

relapsers, an improvement over the 45.4 percent base rate.

76

awn..—

on”:

.32... 3.4.8

Butch ....h «a

.99....6 .. on .m

9... t. a on;

:de—

:Nn..n_

.9339 3...»...

BEU ....n .2.

39:0,. .. cos

powd—

caud—

.99toU .. Saw

.9950 ...on4v

.9utoU 3.de

90.:

.9:

.9350 e on .on

.9350 orwhm

.9250 2.0.;

.cquSEE- A3. x .922
...... x ......2

22.9595 :....33
898.39% . meta—.9...
€959.54. . mate—.9...—

382... 8...... 8:88.36

 

 

 

 

 

 

 

 

 

co. 2. « :..- cc. 3. 9 3.. co. 2. . 2... 39899.99..— 3.8....—
cc. cc. « .5. cc. cc. . cc. 9... 5. fl 5.. 2... cc. . 3.. 2.9.5.3.... E man: .92
a... ..m. 8 9:... co. cm. a on..- so. on. u 2.. I..- an. . 2... 22.6.5... E9639;
-. -- cc. 3.. u :.- cc. 3.. a on: .. .. Lita:
-- : .....- ...n u. .....- .. .. -- -- 2.35.3.5
-- .- -- .- .. .. 8 8. . ... 8:33....
9., n... . m... -- -- -. .. .. -- 9m4.
: z 1 -- I 1 -. -- .993.—
.. -- -- -- l : .. : . 7.9m
x mm a a ~— mm o n x mm a a x mm a g 5.2.5...
3mm ".1: m 9...; Ann". ad. a 95:. 3mm n 3 m 98:. 3.5. u ...... n 9E...
989.9% mg 9853. .2397.
.. ....A ... ... ....o.........-9...-.... ... ... o .o . .4. ...

o. 93....

77

.99toU no... I.

.9950 5...... .m

.359 :93

.859 .....m. :.

.859 5.3

399.5,. o 2..:

23.9595 =29>O
towns—9m - ”5.9595
6953!? . 3:959..—

_....52.... .....54. 5.5556

 

 

 

 

 

 

 

:..: R. a 3.. .- .. -- .. m 3%.”... man.
.- -- -- :2. R. 9 2.. z N 3.53. .232.
NE... 3: 35.55 - m .....m
..N... :8... :...mm 55.2%. A 1. x 3.52
an... :3... :...m... 3.... .352
.359 :18 .359 .. at... .859 .. . . ... .359 .... .8 5.5.3... :55
Bah—DU can .Ow Bah—Orv a och.” 30th «vb-v.6“ Bub-0U o 064.». whats—UN- . wcmuumfiubg
“outcry oowHN Saba/u =9 _ .3». “00:00 orfldh ~99...er owoéh 395331.. - actumfluufi
.95.... .... 5...... 5.32.339
.5. w... H no. 2.. N. on. o... .5. H 2..- :2. 2. 32:5 9523!.
.5. .... 1.. 2.. . :..- .5. .... a .... 2.. 2.. ”8.5.2.... 62
.9..- 2. . an... 14 2. 8. Q. a a... :2. N1.“ 2.. Na: 5.52 5.52
e... .... ... 2.. 2. u 2.. 2.. 2. u 2..- 2. . 2. a 2.- 9.25.5 .9353
a .....3 m. - m AEm
x mm a .. um . .. x ..m a .. x 5.2.5...
32 u 3 m 2..: a; 32 u a. n 3.: 22 u 3 n

 

 

3%.“... .23....

 

78

_o. u. do. 2.. do

.vowasuxu— m5£3£<no 39:01 man and .9324. aficuemkunzx l .030 .253”: 52950 Eon—39:. 6252.."— 02:2:

3.5.5.0 .9ch 3:32 ”coho—gml 9.363 .oZuc 283338..“ ”cactus? 503?."— ozéfiuo 003. ”0121 6.33%: mom ”a. 83st? 39.5.2.0... 305 8.. $.53 .3 5:985 25.

 

.33m 0:39:90 u .QEOU £33.92 u ..92 632:9 :90; 9.5: 3233., __a his 330:. :3 up: 53.. 9.3 312%.... 3:33, :4 .152: 2: 5 v9.35 .0: v.95 932:3 3.39:4:— ..--.. 3:2

 

.00—H—

5.0—6—

.8:5 .1 .2.

.359 93.3

33.50 a i. .3

N 9.5. 3m

.awd

amwd

39:00 91.:

.9230 33.250

82.50 91.:

8 EoEo>an_ A3 ~ x 1692
-- E" x .332
233695 =Eo>O
twang—um - 952qu
m moi 3n £95334. - 9.5595

382.3 5:5. 2:335

 

 

z: a Z... a. ... n: _ l. a 3.. ...... a. 8.. 2225
::. 2.. N m. -- -- .:. no. u a. -- -. 222: E3233
2.. :.. .fi 2.- -- -- 8. 2. a as. .- -- "3.3.3.? .97..
.2. a. .H 2.. -- -- 8. a. a no. .- -- mg; 9.5.42 u.5292
co. m _. g 3. -. -- 8. m _. H 5.. -. .. mcagam .9338
3223., ".2... - w 3a
:32 -. :2: -- ...ueu.,oae_ :: x .252
:33 -- :3: -- 5k .982
x mm « a x ”...,” .. a a ”.5 ¥ 4 x an d s .323...

 

 

 

 

32 n s N. as:

32.. n .3 n 98:.

3mm 4 3 m 98:.

 

 

39:3. .2324

 

79

The logistic models were far better at predicting drug relapsers than abstainers, in
contrast to the models predicting alcohol relapse. For instance, after the first step in both
Time 2 and Time 3 drug relapse, the models could correctly predict who would relapse
with over 95 percent accuracy. Employment status was significantly related to Time 2
alcohol relapse (R = -.l l, p < .01). Individuals with a history of not being employed
were more likely to relapse. In addition, greater psychiatric distress at Time 2 was
related to an increased likelihood of relapse at Time 2 (R = .19, p < .01), just as Time 3
psychiatric distress was related to Time 3 relapse (R = .l l, p < .01). Finally, greater
Time 2 network member substance use was paradoxically related to a decreased
probability of relapse at Time 3. However, when the total efi‘ect of Time 2 Network
Member Substance Use upon drug relapse is considered, this efi‘ect is no longer
significant (discussed in next section). The final model for Time 3 drug relapse predicted
58.7 percent of abstainers, an improvement over the base rate of 39.4 percent, and
predicted 83.3 percent of alcohol relapsers, an improvement over the base rate of 60.6
percent. The results of these regression analyses were used to generate Figures 2 - 5.

One of the great advantages of using path analysis as a data analytic method is
that the path model can be used to generate the indirect effects of the predictor variables
(Ritter, 1988). Although the direct impact (i.e. the beta weight) of one variable upon the
substance use outcomes may be very minor, the indirect effect might be sizable. Thus,
by computing the indirect effects of all of the variables in the four models, the most
accurate accounting of each variables' impact upon the dependent variables can be

assessed. The direct, indirect, and total effects of each independent variable are featured

80

for the ASI variables (Table 11) and for the relapse variables (Table 12). Every possible
indirect effect could not be presented because of the vast number of indirect effects. For
example, there are 16 indirect paths leading from experimental condition to Time 3
substance use and experimental condition is only one of at least 13 variables featured in
the four models. Therefore, only the most important paths are presented in the tables.
The decision rule for including an indirect path in the table was that the magnitude of the
indirect path must be at least equal to .04. This criterion was chosen because indirect
effects are the product of direct paths and the product of two small effect size (.20) paths
(Cohen, 1988) would be .04. However, all of the direct effects and the sum of the
indirect effects are presented for each predictor variable.

Each indirect effect was computed by simply multiplying the path coefficients for
each variable pair along the path. The total indirect effect is the sum of all indirect paths.
For example, the indirect effect of Network Substance Use2 on the Time 3 Alcohol
Composite can be assessed by summing the effect of the six indirect paths. Only one of
the six paths had a sizable effect, the path via ASI Alcohol Composite2. This indirect
path has an effect of .11, which is the product of the two beta weights .20 and .55. The
sum of the other five paths is quite negligible and equals -.02. Finally, the total effect is
the sum of the direct effect (.09) and the sum of all indirect effects ( .09), which equals
.18. This example illustrates that the total effect of a variable can be more substantial
than just the direct effect. The value for the total effect of Network Substance Use2 (.18)
is also similar to its zero-order correlation (r = .20) with ASI Alcohol Composite2.

The total effect (-.12) of family involvement during treatment upon Time 3 ASI

Drug Composite scores is larger than the direct effect (-.07, n.s.) alone. The positive

81
impact of greater family involvement, therefore, is underestimated by examining only its
direct effect. In this case, the indirect effect (i.e. the remaining -.05) was mediated by the
Time 2 ASI Drug Composite. In fact, in every case where the indirect effect exceeded
the .04 cutoff, the indirect pathway, which ended in a Time 3 variable, included that same
variable measured at Time 2. This occurred because the autocorrelation between the
Time 2 and Time 3 measures was typically very high, indicating stability between
measurements. Similarly, the direct effects of Time 2 psychiatric distress upon Time 3
ASI Alcohol and Drug Composite scores (-.08 and -.04, respectively) are both negative
and fail to reach statistical significance. However, the indirect effects of these variables
are quite large. Psychiatric distress at Time 2 is associated with greater Time 2 ASI
scores, which in turn are related to greater Time 3 ASI scores. The total effect of
psychiatric distress measured at the 2-month follow-up is .21 upon 6-month ASI Alcohol
Composite scores and .27 upon 6-month ASI Drug Composite scores.

The indirect effects for the relapse variables are typically quite modest because so
many of the path coefficients were minute or nil (0.00). However, one inconsistency
noted earlier with the regression findings is cleared up when the indirect effects are
computed. It was noted earlier that Network Member Substance Use2 was related to a
reduced likelihood of drug relapse at Time 3 (p = -.09, p < .05). However, once the
indirect effects are combined with this direct effect, the total effect was no longer
significant (total effect = -.02). Therefore, this analysis was able to remedy the only
unexpected and counterintuitive finding, that being surrounded by substance users
actually decreases drug relapse risk prospectively. The results for the psychiatric distress

variable again demonstrate that although the direct effect upon relapse is nonsignificant,

82

the total effect suggests that greater psychiatric distress at the 2-month interview is

related to a greater probability of relapse at six months.

83

daocoe< =8£< u _ .223 u o v38 m. 83. 25362 n .82 How— d=$6=8 5856 m. 38$. 95
9732—8 5:08-~ m. N25... .958 0:89:00 .382 ~m< 95235 .39: .023 BEES. .~ 95$...—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

9.0m.
Mann-QT Nap—ED
oEntham uraniuhnm
o p on . N
h miss: #53327 F Nana—52 «83qu
33.35.... .02 85.55 .3 .02
. ._ _ .-
. . ... h m uw<
QQON om-f 00—.
b «gab. 5U
30:823.: 31... u
m .. u
Masai—:30 A «832—560
_2—OO_< —m< 5 .mm. _2__OU-< ~m<
l H .2539
nob—f
83.30 5253.0
9 . ON . 8:680
fl .ssegm
m8: 02535 233..
84:52 «3.3.92? J 5mg“ {azumw
govv.
gen—4.02 ..o R N 2.82 ..o
m tarpon f 2 am bum—gem
oovv.

 

 

 

080830

330605 moEF

£0668; Noe; owcmzomaéi

2:03:02 n :02 ”Hum d:-30=£ 5:086 fl M08;
0:0 9-30:8 £:0.:-~ m: mofifi 0:000. 0:009:00 mac ~m< wczoaooa .002: .050 BEEF—t .m 0.5!.”—

utfifiuai On . 05020.»!—
0 0 ..

 

 

 

 

 

 

 

 

 

 

 

84

 

 

  

 

 

n35: 3.3.0.? 50:32»: «538$
29550 oz . .2... F 8:53 ..0 .02
i I 0M .
0 0 v. m. .cu—S—ghfl
39:02.3:— 21:0,:
3:39:00 A wanes-:00 fl 553.05.: g
0...: _m< :2. we: .3.

 

 

 

 

 

 

 

fl 8:03 «gnaw
82080
m ......E. p

j ~30 oaaxaw
F

69:02 3.5.502

 

 

 

m8: 8.533
.352 45.202

“omv.

2:02 :0 N 4.02 ..0
m ...:A r 0.0%.:

 

 

 

oymv.

080050 0:20:00:— moect 0:000:02: ~08; 0&058505

85

0003.02 u :02 mam 0.0-32.3 5:08

-0 0. M08... 0:: 00-325. ...:08-~ 0. ~08; 0000.0: .0..00.0 $00.00.... .0008 .0000 0088...... .0 02%.:

    

 

 

 

 

 

 

000m.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3230* R 3230
05020.30 0.0.020an
000m.
. T t p
8.3802 u02.00... N20060:. 0103.02
2.52:0 02 A 3.2.3.... .02
d 3.3m .8102 g
30.52.: SW
gfii. 3.8-“.
380.3. A 3000.0”.
7.3.82 0 0 N M . —O£8—<
HWEEEBQ.EESREWH
.Bfi=aao
N 383893
QSDuESREm NHAHRESiEWH
50802 300302 T f 50802 0.3.5.02
00Nv.
H m 802002 39+ s N 802002w
m 522.00 K 0. am 30.2.00
00mv.
OEOOHSO muOuomfiohn— Mop—:.. manomfioum NOE—P QNHNLOmmonQunm

86

0000.08”... u . 0000.080 .02 u c 00000 0. 80050—005. 2:03.02 u ..02 ”3 00:50:00 0808
-0 m. 38.... 0:0 00-32.00 5:08-~ 0. ~08... 0000.0: wE0 980.020 .0008 .0000 0088:... .m 0.5mE

 

 

 

 

 

 

 

 

000m.
2.0T j Napalm.
0.0.00.9?0 0500.02.00
000M.
N.
«€30.02 0.09502 Nana—=02 3.53.02
.305... 02 F 32...... :0 .02

 

 

   

0BES2QEWH

m ......
000080.02: 2.80...
. ....

3000.00 00.5 H .8.—o

M 830.30 «gunk—w

fl 80680
8550.800

000—.

 

 

 

 

 

 

     
     

 

 

M80200 00.5

 

 

 

 

 

000v

 

 

   

 

 

 

 

 

 

 

 

30 8..-3% ~80 8.538
2.52 «5.202? f 23.52 08.502
0000.
MHUQJU—éuuz no a N 4.02 ..O
mxfitéfi f, awhfitifi
00m0.

 

 

0800000 80.0.0000 $8.... 8000.00.00 N082. 0800.85-20

Table l l

87

II' II’ II lEfli EISI ll :

 

 

 

Outcome Variable
ASI Alcohol ASI Alcohol ASI Drug ASI Drug
Independent Variables Composite2 Composite3 CompositeZ Composite3
Experimental Condition
Direct Effect -.09 .00 .02 .03
Indirect Effect .00 .00 .01
Via ASI Alcohol 2 -- -.05 - -
Via other paths .. .00 -- -
Total Effect -.09 -.05 .02 .04
Treatment Duration
Direct Effect -.02 -. 17 -.07 -.07
Indirect Effect - -.01 - -.03
Total Effect -.02 -. 18 -.07 -.10
Family Involvement
Direct Effect -.06 .08 -. I4 -.07
Indirect Effect -- -.O3 --
Via ASI Dng -- - - -.05
Via other paths - - - .00
Total Effect -.06 .05 -. l4 -. 12
Demographics - Marital
Direct Effect -.08 -.02 -- --
Indirect Effect -- .,o4 -- ..
Total Effect -.08 -.06 -- --
Demographics - Age
Direct Effect -- .10 -- --

Indirect Effect

88

 

 

 

Outcome Variable
ASI Alcohol ASI Alcohol ASI Drug ASI Drug
Independent Variables CompositeZ Composite3 CompositeZ Composite3
Total Effect -- .lO - -
Demographics - Employment
Direct Effect -- - -.O6 ..
Indirect Effect -- -- -- -.02
Total Effect -- - -.06 -.02
Demographics - Sex
Direct Effect -- .05 .- ..
Indirect Effect " " " "
Total Effect -- .05 -- .-
Sobriety Support2
Direct Effect .04 -.05 .10 -.03
Indirect Effect -- .01 --
Via Sobriety Support2 - .- -- .04
Via other paths -- - - .01
Total Effect .04 -.04 .10 .02
Network Substance Use2
Direct Effect .20 .09 -.05 -.07
Indirect Effect -- -- .00
Via ASI Alcohol CompositeZ -- .ll -- ..
Via other paths -- -.02 -— -
Total Effect .20 .18 -.05 -.07
No. of IntimatesZ
Direct Effect -. l l -.03 -.07 .05
Indirect Effect -- -- -.02
Via ASI Alcohol CompositeZ -- -.O6 -- --

89

 

 

 

Outcome Variable
ASI Alcohol ASI Alcohol ASI Drug ASI Drug
Independent Variables CompositeZ Composite3 CompositeZ Composite3
Via other paths -- .03 - -
Total Efl‘ect -. 1 1 -.06 -.O7 .03
Psychiatric DistressZ
Direct Effect .37 -. 12 .34 -.04
Indirect Effect .. ..
Via ASI Composite ScoreZ - .20 -- .13
Via Psychiatric Distress3 -- .13 --~ .18
Via other paths -- .00 -- .00
Total Effect .37 .21 .34 .27

 

Hate, The total direct and indirect effects are featured for all variables in the models. The indirect path is
individually labelled (see row beginning ”Via...") if the path effect 2 .04.
-- Indicates that the path was either not tested or the path effect was <04.

90

Table 12

 

 

 

 

Outcome Variable
Alcohol Alcohol Drug Drug

Independent Variables Relapse2 Relapse3 Relapse2 Relapse3
Experimental Condition

Direct Effect -.04 .00 .00 .00

Indirect Effect .00 -.Ol .00 .00

Total Effect -.04 -.01 .00 .00
Treatment Duration

Direct Effect -.04 -.05 .00 .00

Indirect Effect -- -.01 - .00

Total Effect -.04 -.06 .00 .00
Family lnvolvernent

Direct Effect .00 .00 .00 -.05

Indirect Effect - .00 - .00

Total Effect .00 .00 .00 -.05
Demographics - Education

Direct Effect -- - - .-

Indirect Effect -- .01 .. ..

Total Effect - .01 - -
Dernographics - Employment

Direct Effect .- .. -,1 1 ..

Indirect Effect - -~ -- -.03

Total Effect -- -- -. l 1 -.03

91

 

 

 

Outcome Variable
Alcohol Alcohol Drug Drug
Independent Variables Relapse2 Relapse3 Relapse2 Relapse3
Sobriety Support2
Direct Effect -.02 .06 .00 .00
Indirect Effect -- -.01 - .00
Total Effect -.02 .05 .00 .00
Network Substance Use2
Direct Effect .15 .05 .04 -.09
Indirect Effect .. ..
Via Relapse3 - .05 - -
Via Network Member Substance Use2 -- .00 -- .06
Via other paths -- .00 - .01
Total Effect . 15 .10 .04 -.02
No. of IntimatesZ
Direct Effect -.05 .00 .00 .00
Indirect Effect - -.02 - .00
Total Effect -.05 -.02 .00 .00
Psychiatric DistressZ
Direct Effect .12 .00 .19 .00
Indirect Effect -- ..
Via Relapse2 -- .04 -- .06
Via Psychiatric Distress3 - .06 -- .06
Via other paths -- .00 - .00
Total Effect .12 .10 .19 .12

 

N21:- The total direct and indirect effects are featured for all variables in the models. The indirect path is
individually labelled (see row beginning 'Via...”) if the path effect 2 .04.
- Indicates that the path was either not tested or the path effect was <04.

CHAPTER 4
DISCUSSION

The findings of this investigation affirm previous studies suggesting that social
networks are related to substance use following treatment completion. The magnitude of
the network's impact, however, was quite modest in this study. Furthermore, the most
important network dimension appeared to be the substance use of other members, with
substance use by other members jeopardizing abstinence. Support for maintaining
sobriety and family involvement during treatment, interestingly, were related to only
Time 2 ASI Drug Composite scores. Thus, the negative behaviors of a handfiil of
network members may exert more influence upon resolution of substance abuse problems
than the provision of support provided by concerned relatives and friends. Psychiatric
distress, however, clearly emerged as the strongest predictor of negative outcomes in this
study. The relationship between psychiatric distress and substance use outcomes was
consistent for both alcohol and other drugs and the impact was robust.
Wis:

The social networks of the dually diagnosed participants in this study were quite
small in comparison to other studies of the general population, substance users, and
persons with psychiatric problems. Participants' social networks contained about five

members, on average. This contrasts with prior studies of the general population who

92

93
have around 25 to 50 members and around 40 members on average (Hammer,
Makiesky-Barrow, & Gutwirth, 1978). Among substance users, opoid abusers have
reported to have an average of 12.7 members (Fraser & Hawkins, 1984b) and an average
of 8.5 members for alcoholics in treatment (Gordon & Zrull, 1991). Direct comparisons,
however, between the social network sizes in this study and other studies are not possible
because different measures have been used to identify network members.

Psychiatric populations, similarly, have reduced social networks, with some
indication that these networks shrink with more treatment experience. For instance,
Lipton (1981) found that first admission schizophrenic participants had an average of
15.5 members in contrast to 6.3 members for persons with multiple admissions.
Sokolovosky (1978) discovered that schizophrenics with residual symptoms had
networks containing an average of 10.3 members, but individuals without residual
symptoms had 14.8 members. Holmes-Eber (1990) reported network sizes closest in size
to this investigation - individuals reporting 3.1 average members. Relatives constituted
about 46 percent of the network, similar to the 56 percent figure observed in this study.

Given the small size of the personal networks reported in this study and the
difficult life circumstances facing this dual diagnosis population, it is possible that the
networks' impact may have been muted. In addition, because the networks were so small,
perhaps the impact of any particular member was actually amplified. These hypotheses
cannot be tested in this study, but are worthy of consideration in fixture studies.
Wastes

The discussion of the different model components and hypotheses should begin

with the pre-discharge factors from the model. In general, the impact of these variables

94

(i.e. treatment condition and duration, demographics, family involvement in treatment)
was quite limited. This confirms prior work by Cronkite and Moos (1980) showing that
post-treatment factors appear to be better predictors of outcomes than pre-treatment and
treatment factors. The paths leading from experimental condition to sobriety support and
network member use were not significant. That is, the possibility that the treatment
program may do a better job than the usual care condition of mustering a more supportive
post-discharge environment for participants was not supported. As mentioned earlier,
participants in the experimental condition spent more time in treatment, thereby
confounding treatment condition and treatment duration. Although, treatment condition
was not significantly related to substance use outcomes, treatment duration was.

Longer treatment duration was related to fewer alcohol problems at the 2-month
follow-up. This supports earlier studies (Goehl, Nunes, Quitkin, & Hilton, 1993;
Simpson & Sells, 1983) that longer treatment duration is related to more positive
outcomes. The role of treatment duration appears, however, to be quite modest when one
considers that there were eight opportunities (or direct paths) for significant relationships
with the outcome variables. For example, there were four outcomes (alcohol relapse,
drug relapse, alcohol problems, and drug problems), all of which were measured at two
different follow-up periods (2- and 6-months post-discharge).

Similarly, family involvement during treatment appeared to be another modest
predictor. Greater family involvement was associated with fewer drug problems at the 6-
month follow-up. It should be noted that this measure was the only predictor variable
coded from existing records; all other variables were based on the self-report of the

respondent. The advantages of using different data sources is obvious and observing a

95

significant predictive relationship with this measure would have helped establish the role
of family involvement in affecting outcome. Limitations in the way this variable was
conceptualized are evident - only one item was used to assess this construct and hospital
stafi‘ did not always know how much support was provided by family and how significant
that support was. One psychologist who had worked at Northville Regional Psychiatric
Hospital mentioned that stafi~ did not always know when family members visited patients
because the staff were on the wards during particular hours or days.

5.1” lSl'S lEl"D'

It was hypothesized that the social network and support variables would be
related to improved outcomes and the impact of having network members who abused
substances was hypothesized to have the greatest impact. Sobriety support was quite
limited in guarding against relapse or problem substance use. Of all eight Opportunities
for significant relationships with outcomes, only one significant association was
observed. It may be possible that the method of assessing this construct may have
hindered its impact. Although only one question was used to assess this construct, it was
asked for each network member. Therefore, the average individual would have
approximately five ratings for this measure. Prior studies in which both support and
network measures have been administered, typically have shown that network member
use is a better predictor of outcome. Thus, the findings observed here confirm prior work
in non-dually diagnosed populations and support the hypothesis that sobriety support
may be less influential than network member use.

Of the three social factors assessed (i.e. sobriety support, network substance use,

number of intimates), network member substance was most strongly tied to outcomes as

96

hypothesized. For instance, individuals reporting greater network member substance use
at Time 2 were more likely to relapse using alcohol (Time 2) and had higher Time 2 ASI
Alcohol Composite scores. Time 3 Network Member Substance Use was related to an
increased probability of drug relapse. However, greater Time 2 network substance use
was also associated with a lower likelihood of drug relapse at Time 3, a counterintuitive
finding that no longer held once the indirect effects analyses were conducted. Overall,
the direct efi’ects of this social network measure upon either cross-sectional (2-month) or
longitudinal (6-month) substance use measures were quite nominal. In contrast, the
indirect effects of network member substance use suggested that this construct is a viable
prospective predictor of outcome. These findings are not necessarily unexpected. For
instance, Tausig (1992) noted that when network structural properties are related to
health outcomes, small or insignificant effects are likely. However, it is often the indirect
efi‘ects through mediating variables that are ultimately more important.

The number of intimate members, or confidants, was only related to Time 2 ASI
Alcohol Composite scores. Having more confidants was associated with fewer alcohol
problems. The number of confidants may be more strongly related to the psychiatric
measures than to substance use. At the 2-month follow-up the zero-order correlation
between the number of confidants and psychiatric distress was significant, indicating that
lower distress was associated having more confidants. Although, at the 6-month follow-
up, the correlation was not statistically significant.

In summary, these network and support measures were modestly related to the
outcomes. Other studies that have observed more robust findings typically did not assess

(and control) for as many extraneous variables. For instance, demographic variables or

97

prior substance use were not controlled (i.e. statistically) in any of the other studies cited.
The other explanation for the modest impact of the social network characteristics may be
that the sparse social networks for this dual diagnosis population may only modestly
affect substance use.

Psychiatric distress, as measured by a combination of the SCL-lO and the ASI
Psychiatric Composite score, demonstrated strong cross-sectional relationships with the
substance use outcomes. In fact, all eight of the eight cross-sectional paths were
statistically significant. Furthermore, the magnitude of the coefficients was notable for
the ASI composite scores, with betas ranging from .26 to .37. Prior studies have shown
that severity of psychiatric problems is related to substance outcomes (McLellan,
Luborsky, Woody, & Druley, 1983; Powell, et al., 1992; Rounsaville, Dolinsky, Babor,
& Meyer, 1987). However, the role of specific diagnoses in affecting outcomes is
unclear. Powell et a1. (1992) found that specific diagnoses were not related to substance
use outcomes, but Rounsaville, Dolinsky, Babor, and Meyer (1987) did. Psychiatric
diagnoses were not used in the present study because the hospital diagnoses were not
obtained using standardized assessment instruments.

Perhaps such strong relationships with outcomes were observed because the
psychiatric distress variable was created by forming a composite of two variables with
excellent psychometric properties. The SCL-IO has excellent internal consistency and
both the SCL-lO and ASI Psychiatric Composite score had strong evidence for their
construct validity. However, these effects could also be the result of method variance
since both constructs were based upon participant self-report. Moreover, these two

constructs could be assessing the same underlying construct. This latter proposition,

98
however, seems unlikely because these variables do not show strong external parallelism.
That is, the pattern of correlations between study variables and the psychiatric distress
measures differs from the pattern between study variables and the alcohol and drug
outcomes.

These consistent cross-sectional relationships, however, do not prove that greater
psychiatric distress causes more substance use problems. The reverse may also be true.
There is widespread speculation in the dual diagnosis literature about which is the
"primary" of the dual problems. Dackis and Gold (1992) for instance, assert that
psychiatric problems are more likely to result from addiction. It is unknown whether
some individuals are attempting to self-medicate their psychiatric problems with drugs or
if both problems are caused by the same underlying mechanism (Brown, Ridgely, Pepper,
Levine, & Ryglewicz, 1989). The link between substance use causing psychiatric
distress was not tested in the present study; however, the question is certainly worthy of
firrther study. Kessler and Price (1993) noted, for instance, that the primary disorder may
create a context for the secondary disorder to develop. These authors noted that chronic
depression may lead to the erosion of social networks thereby increasing the risk of
numerous other conditions.

One possibility tested by the model was that substance use may shape the content
and composition of the social network. This idea was tested with the three paths leading
from substance use to (a) sobriety support, (b) network member substance use, and (c)
number of intimate members. None of these three paths were significant in any of the
four models tested (12 total tests), thereby disconfirrning this hypothesis. These findings

among a treatment population confirm earlier longitudinal research on drug use among

99

adolescents. It appears that drug use is more influenced by fiiends introducing substance
use to each other than by individuals actively seeking fiiends who share common beliefs
and behaviors (Kandel, 1978). Finally, it should be noted that the autocorrelations of the
variables across the two time frames were assessed along with the substance use
variables. These autocorrelations were typically quite high (0's in the .40's and .50's)
accounting for much of the variance in the social support/network variables.

Although many of the methodological limitations of prior studies were remedied
in this investigation, this study still has several limitations. All of the variables assessed
in this study were based on participant self-report with the exception of the family
involvement in treatment variable. The accuracy of self-report network data has been
questioned by some (Fraser & Hawkins, 1984a). However, a review by Marsden (1990)
has concluded that the types of network measurements that were used in the current study
have demonstrated adequate reliability and validity. The network measures were all
administered to persons receiving treatment for some form of mental illness, a group that
some would argue might have difficulty providing accurate responses. Although there is
more data on the validity of social network measures for the general population, some
studies have been conducted with psychiatric patients. According to a study of 29 clients
with psychiatric disorders by Moxley (1988), core membership and social network
structure can be reliably reported. Attempting to establishing the validity of social
network data, however, is often difficult because other informants have been shown to
underestimate a person's network because they do not know all of the focal person's

acquaintances.

100

The measures of substance use in this study, like the social network measures,
may have suffered from some form of self-report bias. Biochemical verification
techniques such as hair, urine, and blood testing are available, but were not used in the
present study. On the positive side, the substance use measures based on the Addiction
Severity Index are well validated and have been shown to be related to independent
indicators of substance use problems. Standardized substance use instruments, however,
have been shown to have poor sensitivity and specificity when used with schizophrenic
patients (Drake, et al., 1990). Thus these authors urge other researchers to supplement
self-report and interview data with information obtained using behavioral observations
and through reports from collaterals and case managers.

In addition, there is some controversy over using a simple dichotomous relapse
coding scheme. Marlatt (1985b) has criticized this dichotomous approach where relapse
is seen as "an enigma: the end of the road; a dead end” (p. 31, italics in original). This
approach, argues Marlatt, may set up a self-fulfilling prophesy for patients who may
believe that any return to substance use is a violation of abstinence and could then lead to
a full-blown relapse. Moreover, this dichotomous view of relapse suggests that there is a
return to the disease state, which Marlatt felt is not very accurate.

The social network measures used for this study were based on the most basic
social network indicators, such as the number of intimate members, the substance use of
members, and the provision of sobriety support. Several morphological network
characteristics, such as network density, reciprocity, and linkages, were not measured in
this study. These morphological network measurements can help elucidate the processes

by which networks influence the resolution of substance use problems or their escalation.

101

F urtherrnore, the study results are generalizable only to an urban, predominantly
lower socioeconomic status population. Compared to other dual diagnosis investigations,
the study population. has more Afiican American participants and fewer schizophrenics
(Kay, Kalathara, & Meiner, 1989; Lyons & McGovem, 1989).

EutnrcResearch

Future research in this area should explore the events surrounding the relapse
episodes to understand how network members facilitated this process. Also, nearly half
of the participants had abstained from alcohol or other drug use prior to their follow-up
interview -- what factors influenced these individuals to maintain their sobriety? A
detailed accounting of the role each network member played in either relapse or
abstinence would greatly add to our understanding of the relapse process. Asking the
length of time that the participant knew each network member can be usefirl to see if the
network membership is evolving. The change in the network could be assumed if the
average time the focal person knew network members decreased as new members are
absorbed into the network.

Psychiatric distress was also a strong predictor of relapse and should be studied
further. Prior studies have shown that negative emotional states are related to relapse
(Marlatt, 1985c), thus future research should clarify these factors in light of the co-
existing psychiatric problems facing dually diagnosed individuals. Assessing substance
use outcomes with urine samples or hair samples may help alleviate concerns that these
findings were solely based on method variance.

Future studies might assess the exact date of the first relapse (or lapse) following

treatment. By knowing the number of days until this event occurred (or did not occur),

102

sophisticated techniques such as survival analysis or random regression can be used to
generate relapse curves for individuals with different network configurations. Also, these
analyses might be conducted using different subgroups of substance users to test whether
alcohol abusers might differ from crack cocaine abusers in how networks influence their
substance use.
Conclusions

The results of the current study suggest that psychiatric distress and contact with
substance using network members can subvert attempts at recovery from substance use
problems. The negative impact of contact with substance using members was greater
than the positive impact of greater family involvement during treatment and sobriety
support provided by network members. Therefore, as part of aflercare it is critical to
stress to patients that they must avoid former substance-using contacts. Traditionally,
however, most treatment professionals have focused solely upon intrapersonal issues
while neglecting how interpersonal issues can affect relapse (Smith, F rawley, & Howard,
1991). In addition, the dually diagnosed individuals must actively monitor their
emotional health once discharged and seek assistance from family, friends or treatment
professionals as needed. This recommendation parallels the wisdom of self-help
organizations, whereby addicts are warned that relapse can be caused by HALT -
becoming Hungry, Angry, Lonely, or Tired (Narcotics Anonymous, 1988). Dealing with
emotional states has also been incorporated into relapse prevention programs, which
acknowledge that changes in mood, attitudes, and thoughts are all warning signs for

relapse (Daley & Marlatt, 1992).

 

103

One of the most encouraging aspects of conducting social network research is
knowing that social networks can be modified -- networks are not immutable
characteristics such as demographics or genetics. For helping substance abusing
populations, several interventions, often termed network therapy (Galanter, 1987;
Galanter, 1993; Sorensen & Gibson, 1983) are available to shape the social network and
work with network members to facilitate the recovery of the focal person. Hawkins and
Catalano (1985) noted that the first efforts to assist families and network members of
addicted persons began in the early 1970's and have since grown in number and
influence. Many of these efforts have been based on self-help group models, such as
Alcoholics Anonymous, Narcotics Anonymous, or Al-Anon. Also, in the area of mental
health, interventions involving patient's social network have already been developed and
shown encouraging evidence for their effectiveness (Kleiner & Drews, 1992; Schoenfeld,
Halevy, Hemley-van der Velden, & Ruhf, 1986 ). Moreover, assertive case management
has been shown to improve social functioning and increase network size for homeless
mentally ill persons (Wasylenki, Goering, Lemire, Lindsey, & Lancee, 1993).

In conclusion, the social network and psychiatric measures accounted for a
sizable amount of the variance in post-treatment firnctioning. These findings may be
usefirl for developing specific interventions for dually diagnosed persons, who often have
poorer prognoses than individuals with either psychiatric or substance abuse problems.
Furthermore, the network enumeration methodology employed in the present study may
be useful for other studies involving dual diagnosis populations or persons with substance

abuse problems. Further work should investigate the psychometric properties of the

 

 

104

instrument and other researchers may choose to incorporate more detailed network

questions.

 

 

APPENDICES

APPENDIX A

SAMI Project Participant Consent Form

 

105

APPENDIX A

SAMI Project Participant Consent F orm

MICHIGAN STATE UNIVERSITY
SAMI RESEARCH PROJECT
PARTICIPANT CONSENT FORM

PURPOSE

There are two main purposes of the SAMI research project. The first is to collect
information about the Chemical Dependency Unit at Northville Regional Psychiatric
Hospital and the second is to find out what happens to people after they leave the
hospital. We will be contacting and interviewing participants periodically for up to 2
years after discharge from Northville.

PROCEDURES

Participating in this research study involves the following:

Amnuntnfficntactflithlnu;

We would like to interview you several times throughout the next 2 years. In the
hospital we will be asking you for 3 interviews: right after you consent to participate,
after you have been in the hospital for a short while, and right before you are discharged.
These interviews will take approximately 1 hour to complete.

Once you leave Northville, we want to interview you every 4 months for a total of 5
times. These interviews will take approximately 2 hours to complete. Together we will
choose where we will conduct the interviews so it will be convenient for you. Also, we
would like to be in touch with you by mail or telephone every 2 months just to make sure
we can locate you for your next interview.

WW

By agreeing to participate in the project, you are also agreeing to accept one of two
possible firture ward assignments. The alternatives are assignment to the Chemical
Dependency Unit (CDU, A-S-l) or remaining on the ward to which you are currently
assigned or another ward to which the hospital would normally assign you. This ward
assignment will continue throughout the course of the study. The choice of these
alternatives will be determined solely by lottery. This means that one-half of the people
in this study will be in the CDU and the other half will be assigned the way the hospital
typically assigns people to wards. The choice of treatment unit assignment in this study
is strictly random.

106

mm

During interviews with you we will ask a lot of questions. Your answers will be kept
strictly confidential. We will want to know about many aspects of your life. Specifically
we will ask about your alcohol and drug use, your physical and emotional health, your
social relationships and the kinds of activities you do, your job, legal problems, your
living situation, your attitudes about substance use, your goals in life, and the mental
health or substance abuse services you have used.

Additionally, while you are in the hospital we will want to know how you felt about
the services your received at the hospital, which activities you participated in while on
the ward(s), and what your plans are for the future. We will also ask you questions about
your knowledge of substance abuse and mental health problems.

W

We will ask you to give us permission to contact a few people you know and
agencies you may be receiving services from to ask them if they know how to locate you
if we cannot find you. We will tell them that you are participating in a research project
concerning your treatment at Northville Regional Psychiatric Hospital. We will only ask
them to help us contact you. We will not ask fiiends, family, or significant others any
other questions.

I . . C] l .
We would also like to obtain information from staff at agencies you receive treatment
from after you leave Northville Regional Psychiatric Hospital. In addition, we would
like to have access to your clinical records and management information systems data at
Detroit/Wayne County Community Mental Health Board and its contractual service
providers for the purposes of obtaining information on 1) any information that may aid
us in locating you, 2) access to the dates, the number of times, and the types of services
you have used, and 3) answering questions regarding your community adaptation.
During the various interviews that we have with you, we will ask you for names of
agencies that have provided services to you from community mental health, substance
abuse, or other service agencies. We will then contact one or more of them to ask if they
will answer some general questions about how well things are going for you. They do
not have to participate if they do not want to be involved.

Benefitslalcu;

In the past, many people have found participating in this type of study an interesting
and educational experience. For your participation in the study you will receive
monetary compensation for each interview. For completing the 3 interviews while you
are in the hospital, you will receive $15. For the 2 hour interviews in the community you
will receive $10 for each. If you complete all of the interviews you will receive a bonus
of $1 5, which makes the total you can receive $80. Additionally, you will be asked to
send us updated addresses and will receive small gift coupons as compensation.

107

Risks;

We anticipate no risks to you from participating in this study. Some of the questions
we will ask you may be about difficult or emotional subjects. If you are feeling uneasy
about any of the questions, please tell the interviewer that you choose not to talk about
that subject.

III 2....

Your participation in this study is completely voluntary. Whether you agree to
participate will have no effect on the services you will receive from the Department of
Mental Health. You are free to withdraw from participating at any time. You do not
have to respond to any questions you do not want to answer.

All information you give to us during an interview will be kept completely
confidential. Instead of using your name, we will use a code number to identify your
responses. The only people who will have access to your answers will be the research
staff. We will be interviewing approximately 450 people and the data will be compiled
and presented together, not on an individual basis.

Your identity will be protected under our Confidentiality Certificate issued by the
Department of Health and Human Services. The persons authorized by the
Confidentiality Certificate to protect the identity of research subjects may not be
compelled to identify research subjects in any civil, criminal, administrative, legislative,
or other proceeding whether Federal, State or local. If either of the following conditions
exist the Confidentiality Certificate does not authorize refusal to reveal identifying
information conceming research subjects: 1) The subject consents in writing to disclosure
to identifying information, or 2) authorized personnel of DHEW request identifying
information for audit or program evaluation of a research project funded by DHEW or for
investigation of DHEW grantees or contractors and their employees or agents carrying
out such a project. The Confidentiality Certificate does not govern the voluntary
disclosure of identifying characteristics of research subjects. The Confidentiality
Certificate does not represent an endorsement of the research project by the Secretary.

QuesticnmLCnncems;

If you have any questions or concerns regarding this project, please call Ms. Marjorie
Ebejer, at (313) 349-1800 ext. 2895, Dr. Douglas Luke, at (517)353-0871, or Dr. Carol T.
Mowbray at (313) 577-8806.

 

CONSENT STATEMENT
I understand all of the information written on this form. I have had an opportunity to
raise questions and have them answered. By signing this consent form I am agreeing to
participate in the study under the conditions listed above. A copy of this form will be
provided to me.

 

Participant Name (PRINT)

 

 

Participant Signature Date Interviewer Signature

APPENDIX B

Social Network Measure

108

APPENDIX B

Social Network Measure

Social Network Section - Instructions

1. Begin by asking probe question #1 at the bottom of the page. For each name that the
respondent mentions, turn to the next page and write down their first name and first letter of
their last name. Continue recording responses for this question until the respondent is
finished and prompt the respondent by asking, "Is there anyone else...?".

2. Next, for the first column of the table, "Probe Number," record the number of the question
you have been asking (In this first instance you would write "1" for Question 1). If the same
individual is mentioned twice, then DO NOT write their name down again, but write down
that question number in the Probe Number column next to where that person's name was first
mentioned.

3. Continue asking questions 2 - 5. When you have recorded all of the names of the individuals
for the remaining questions and the probe number for each one, then proceed to fill in the
remaining columns (e.g. Sex, Know from Northville, etc). For each name go across the row
asking each of the questions for that one individual. Do this until the chart is completely
filled in for each person.

Read:

For this next section. I will be asking you questions about your fi'iendr and people who you are
close to. I will be asking you to list some of the names of your friends; however, I will not ask
you their fitll name, just their first name and the first letter of their last name. This way you can
protect their identity. Do you have any questions?

Probe Questions:
(I) When you are concerned about a personal matter-for example. something you are worried
about or you are concerned about someone you are close to-who do you talk with?
<PROMPT-for each question>
(2) Who do you spend your time with, that is - who do you hang out with?
(3) Who would you ask if you needed to borrow some money?

(4) Do you attend any support groups such as AA/NA meetings?

NO—---> {IF NO, THEN SKIP T0 #5)
YES----> {IF YES, THEN Who are your good friends in this group?)

(5 ) Is there anyone else important in your life who you have not mentioned? <PROBE>

_‘

 

109

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

< D U Q u H- O 3: - " 5‘ -l 2 Z O

 

 

 

 

 

I
a
0050...... .300 5.5.582...
..a . 3.5". a]; feel
:..-56:... «3.9.54 ...0 5:31.
.5 o...... <n~ ...... 5.5-. 3.5.8:...
.1 .. .oz _ 0.8.0?" 02.23.". .3550". 1.31.090...
.821. .0. . 530-. 3 . 3.5.". ...-on. 33.50..
0:... .5. 03582.. 023152.. 583... 92A .3833... a...
8.. 91.! ... 3.5 .... ...... (IN .3 ...... <1. 05.52... 835..-. If. 80 .28 3." 2.. 05:... a...
S. 2.0 ll 8 a .29.... ... a 321 ..- I 82:. 30.31. 2.5.: 2.0.1.... 0.2!. a. 2..: Ice
.80 0500... ....- E €00.56 «303.80 5?. 5.5 ...—9. 3 hi 5... .0808 .- ”Bee... ..eeCk. :5 35.0 no. .05. Eu 0 2:10 .2502
.0355: .80 . u. .. . r 32.55. 2.3020533? 203.25 E03313 32..
2:... 8:0 820 820 8:» 800 3:» 0.20 8:0 .3.“ .oz

 

 

 

 

 

 

 

 

 

 

 

 

APPENDICES C-l to 04

Four Tables of Standardized Path Coefficients

 

110

 

 

 

 

Table C - l
:.t: 05;. 3 : it .t s r‘ '.tust' ' «Neutrino. ‘d‘O\ H».
AlmhnLCompnsiteScnrcs
Follow-up Period
Time2(n=372) Time3 (n=299)
ASI
SM Network Alcohol SW NW BM ASIAW
Predictor Supponz ManhuUsez Comp) SupportJ Manta-rues No. cram-tea DistreuJ Comps
m
5...,- .. .. .. .. .. _ - .05
W .. .. .. .. .. .. - -
A, .. - .. - - - - 10°
M - - - - - - - -
W - - - - ~ - - ~
W .. .. .. .. .- .. -.02
MW 93 -.02 - .. .- .. no
No. Days sum — - ..oz - — - - ..rr-
Furiy Imam - - -06 .. - - .. .os
W
WW - - .04 .44" - - - ..os
Ndwol'kM-nchuZ - - .20" - .44" - - .09
Mortimer-2 - - -.ll’ - - .36“ - -.03
WW - - .37" - - - .50" -.|2‘
mm
5mm - - ~ - - - - --03
Network Matthew - .. - .- -. — -.05
No. cram - « - .- ~ - - .07
w Dine-s3 - - - -. - -. — .26“
W
asrmcm: - - - .0: .oo .02 — .55“
Alcohol Rehpae! - - — — - - - -
ASI meme: - .. .. - .. - .. ..
WW - - - - - - - -

 

lime. "-" Indicates path was not hypothesized. Net. =- Network; No. - Number; Comp. - Composite Score.
'The direction of coding for these dichotomous variables is: Sex O-Female, l-Male; Race O-White I-African
American: Employment O‘Not working 1=Employed; Marital 0=Not currently married 1=Married; Treatment
Condition O-Usual care l-Experimental; Alcohol and Drug Relapse 0=Abstaining 1=Relapsed.

‘n < .05 ”p < .01

111

TableC-Z
:ttzt 610:“: O’llfl't‘ 0 t‘ ith" .91....W0101 ' ‘I\ I".

W

 

Follow-up Period

 

 

 

Time 2 (n = 324) Time 3 (n = 319)

Predictor m 2 M53130 2810:? m "was No. «m3 m an?
m
Sq' .. .. .. .. - .. - ..
m .. .. .. .. - .. .. ..
A, .. .. .. - .. .. .. ..
see-sou - ~ - - — - .- -.
W - - .06 .. .. .. _ ..
m .. .. .. .. .. .. .. ..
m CM' 03 -.O2 .02 - - - - .03
No. Days'lt'l‘m .. - ..or - .. .. - ..or
ruym - .. ...... — .. - .. ..01
mm
mm - — w .43" — - .. .0:
WWW! ~ - ..os — .0" .. - ..or
worm-.2 .— - or - - as” - or
WW .. - .34" — - - so» ..or
W
memo - - - - - - - or
NetworkMemchsd .. — -- — - - - .os
Ned‘m - - - — - - .. 03
Worm-o .. .. .. .. .. .. ., as"
W
ASlAlcclrchcmpJ - .. .. .. .. .. _
wad-pa -. — - - — - — .-
ASI mom; — - .. or 00 oz .. 31»
MW - - - - - - - -

 

Nate. "-" Indicates path was not hypothesized. Net. - Network; No. - Number; Comp. - Composite Score.
'The direction of coding for these dichotomous variables is: Sex 0-Female, 1=Male', Race 0-White l-Afiican
American: Employment 0=Not working l-Employed', Marital O-Not currently married l-Married; Treatment
Condition 0=Usual care 1=Experimental; Alcohol and Drug Relapse 0-Abstaining 1=Relapsed.

‘p < .05 ”p < .01

112

Table C - 3

5
e

   

 

Follow-up Period

 

 

 

 

Time 2 (n = 374) Time 3 (n = 334)

PrediCtOT Support 2 ~43an 2m 3m urine; No. cram m m
m
s“. .. .. - .. .. _ .. -
W .. .. .. .. _ .. .. ..
A, .. .. .. .. .. - .. ..
M - - 02 - - - - -.
W - - - - - - - -
M-rt-r- - -. - - - - - oo
nan-rt cm 03 02 .. .. .. .. no
No. buys-11mm ~ - - - - - or
Funny swarm-n - - .. .. .. .. on
W
scanty sumo - - oz 43" - - - 00
NM m Uad - - .15“ - .42” - - .05
No. cranium - - ..os - - .35" .. .oo
worsen-2 - — .12“ - - -. so" .00
max-au-
somamo - - - - - - - 0°
mums-was .. .. .. .. - .. .. 00
No. arm-m -. - .. .. .. .. 00
W Dinners) ~ - - - - ~ .- .r l“
W
as: AboholCcmpJ .- -. .. .. .. .. .. ..
About m - - — ..03 .06 .m - .32“
A81 moms - .. - .. - .. .. ..
mum-2 - - - - - - - -

 

N915. "-" Indicates path was not hypothesized. Net. - Network; No. - Number, Comp. - Composite Score.
‘The direction of coding for these dichotomous variables is: Sex 0=Female, 1=Male, Race 0"White 1=African
American; Employment O'Not working l-Employed; Marital 0-Not currently married 1-Married; Treatment
Condition 0=Usual care 1=Experimental; Alcohol and Drug Relapse 0-Abstaining 1=Relapsed.

’p < .05 ”n < .01

 

Table C - 4

113

E 10 l' fllElllllEl" E El

 

Follow-up Period

 

Time 2 (n = 372)

Time 3 (n = 336)

 

 

Predictor my. Mm. 0'2. 0% um mm... m 0......
W
Sex‘ .. .. .. _. .. .. .. ..
m .. .. .. .. .. .. .. ..
A, - .. - .. .. - - .os
Educdcn .. - - - - - - -
Employment - - ..r r" - .. .. .. ..
MIN - - .00 - - .. .. ..
Tm Condition‘ 03 - 02 .oo .. .. .. ..
No. Days in Treatrnau - - oo .. .. .. ..
Family Involvement .. - .oo - - - - ..05
W
Souurysuppora - - .oo .43“ - - .- .00
Network Member Use2 - - .44" - - -.09‘
No. “m2 - - - - .36" - .oo
Prychirtric Dane-r2 - - .19" — - - .50“ .00
W
Sch-Iv w ~ - - - " ' " -°°
mamas: .. .. - - .. _ _ 43..
No. of 1rrtr'matas3 - ~ -. .. .. .. .. .00
W Distress3 .- - - - - - - .r r"
W
ASlAhoholCcmpJ - — .. .. .. - .. -
Alcohols-spa - - - - - - - -
ASI Dru. Comp! .. ~ — .. .. .. .. ..
Dru; Ilehpuz - -. — ..02 .01 .oo - .31”

 

m. "0" Indicates path was not hypothesized. Net. - Network; No. - Number, Comp. - Composite Score.
'The direction of coding for these dichotomous variables is: Sex 0-Female, l‘Male; Race 0-White I'African
American; Employment 0-Not working l-Employed; Marital 0-Not currently married l-Married; Treatment
Condition 0=Usual care 1=Experimental; Alcohol and Drug Relapse 0=Abstaining 1=Relapsed.

‘p < .05 ”p < .01

LIST OF REFERENCES

 

 

LIST OF REFERENCES

Akers, R. L. Krohn, M. D., Lanza- Kaduce, L., & Radosevich, M. (1979). Social learning

and deviant behavior: A specific test of a general theory W
Mm, 551, 636- 655

Bachrach, L. (1986-1987). The context of care for the chronic mental patient with

substance abuse problems. BMW, 58(1), 3-14.

Berkman, L. F. (1986). Social networks, support, and health: Taking the next forward.
Americanlnumamffinidemiclcgy 123.. 559- 562

Billings, A G., & Moos, R. H. (1983). Psychosocial processes of recovery among
alcoholics and their families: Implications for clinicians and program evaluators.

Addictimfiehaxinrs 8 205- 218

Booth, B. M., Russell, D. W., Soucek, S., & Laughlin, P. R. (1992). Social support and

outcome of alcoholism treatment: An exploratory analysis. Amerimmurnamflnmg
andAlthnLAhuse. 18. 87-101.

Boyd, C J, & Mieczkowski, T. (1990). Drug use, health, family and social support in
"crack" cocaine users. Addictixefiehayiors, 15, 481-485.

Brown, S. A., Vik, P. W., & Creamer, V. A. (1989). Characteristics of relapse following
adolescent substance abuse treatment. MW, 14, 291-300.

Brown, V. B., Ridgely, M. 8., Pepper, B., Levine, I. S., & Ryglewicz, H. (1989). The
dual crisis: Mental illness and substance abuse. Whom, 451, 565-569.

Cohen, C. 1., & Kochanowicz, N. (1989). Schizophrenia and social network patterns: A

survey of black inner-city outpatients. CommuninLMemalfleathQumal, 2,5,
197- 207

Cohen, E. H. ,Mowbray, C. T., Bybee, D., Yeich, S. ,Ribisl, K, & Freddolino, P P.
(1993). Tracking and follow-up methods for research on homelessness. mutation

Rem, 11, 331 -352
Cohen. J. (1988). StatisticameaeLanalxsiszLthejehavicralsciences (2nd ed).
Hillsdale, NJ: Lawrence Earlbaum.

114

115

Cohen. J. & Cohen, R (1933) Anclrcdmultrnlcregncssrcnfccrrclatmanahtsufcuhc
hchayimaLscicnccs. (2nd ed..) Hillsdale, NJ: Earlbaum.

Coppotelli, H. C. & Orleans, C. T. (1985). Partner support and other determinants of

smoking cessation maintenance among women. IcumaLgLCQnsnltingandflinical
Emhnlagy 53.. 455-460

Cronkite, R. C. ,& Moos, R. H. (1980). Determinants of posttreatment functioning of

alcoholic patients: A conceptual framework. ,[Qumammcnsuhmgandflmcal
Psychology, 48, 305- 316

Curry, S, Thompson, B, Sexton, M. & Omenn, G. S. (1989). Psychosocial predictors of
outcome in a worksite smoking cessation program. AmericanlnumaLQEPchcmiyc
Medicine iLL) 2- 7

Dackis, C. A., & Gold, M. S. (1992). Psychiatric hospitals for treatment of dual
diagnosis. InJ. H. Lowinson, P Ruiz, & R B. Mrllman (Ed) SubstanccAhum

Cgmpnchcnsiyclcxthmflpp 467-485). Baltimore. Williams& Wilkins.

Daley, D. C., & Marlatt, G. A. (1992). Relapse prevention: Cognitive and behavioral
interventions. In J. H. Lowinson, P. Ruiz, R. B. Millman, & J. G. Langrod (Ed),

WWW 533- 542). Baltimore, MD:
Williams & Wilkins.

Davidson, W. S, Redner, R., Amdur, R L., & Mitchel, C. M. (1990).A1icma1i1c

.-NI 0 ot-tott‘.'0t'orutt' ‘ u.New
York: Plenum.

Derogatis, L., Lipman, R., & Covi, L. (1973). The SCL-90: An outpatient rating scale
(Preliminary report). WM 2, 13-29.

Drake, R E, Osher, F C, Noordsy, D L., Hurlbut, S. C., Teague, G. B. ,& Beaudett, M.
S. (1990). Diagnosis of alcohol use disorders rn schizophrenia. Schizophrcnia

Bullctinmlfi5767

El- Guebaly, N. (1990) Substance abuse and mental disorders. The dual diagnoses
concept. Canadranlcumflcffsxchrm, 35, 261 -.267

E11, K. (1984) Social networks, social support, and health status: A review. Social
SmiccRcyicw 5.8,133-143

Ellickson, P. L., & Hays, R. D. (1992). On becoming involved with drugs: Modeling
adolescent drug use over time. Hcahnflsychglcgy, .11, 377-385.

116

Everingham, S. S., & Rydell, C. P. (1994). Madelingjhcjcmandflmmaine. Santa
Monica, CA: Rand.

Favazza, A. R., & Thompson, J. J. (1984). Social networks of alcoholics Some early

findings Alcahclismflinicalandflxncrimcntalltcscarch. 8 9-15.

Fraser, M., & Hawkins, J. D. (1984a). Social network analysis and drug misuse, Sccial
WW0]. March, pp. 81-97).

Fraser, M. W. ,& Hawkins, J. D. (1984b). The social networks of opioid abusers. Ihc
IntcmaticiiaLlcumalcfithcAddicticns. 12, 903- 917

Galanter, M. (1987). Social network therapy for cocaine dependence. Adyanccsin
AlccthandSulzstanceAlzusc. 6. 159-175.

Galanter, M. (1993).Ncmcrk1hcrapitchalccthmddrug_almsc_Ancmnpmachjn
practice. New York, NW: BasicBooks.

Goehl, L., Nunes, E, Quitkin, F ., & Hilton, 1. (1993). Social networks and Methadone
treatment outcome: The costs and benefits of social ties. Ammcanlcurnalflmg
andAlthQLAlzusc. .12. 251-262.

Goldstein, P. J, Abbott, W. ,Paige, W. Sobel, 1., & Soto, F. (1977). Tracking procedures

in follow-up studies of drug abusers. AmmcarrlcurnalgfmaniAIcchQLAbusc,
5(1), 21 3-0.

Gordon, A. J ., & Zrull, M. (1991). Social networks and recovery: One year after inpatient
treatment. lcumaLcflSuhstanccAbusclrcatmcnt. 8. 143-152.

Gottheil, E, Thornton, C. C. Skoloda, T. E., & Alterman, A I. (1982). Follow-up of

abstinent and non-abstinent alcoholics. W012,
560-565.

Gottlieb, B. H. (1985). Social networks and social support: An overview of research,

practice, and policy implications. Hcahhfiducaricnfluancrly, 12, 5- 22.

Grey, 0, Osborn, E., & Reznikoff, M. (1986). Psychosocial factors rn outcome in two
opiate addiction treatments. InurnaLgfflinicaLEsycthcgy, $1.2, 185-189.

Hammer, M. (1983). Core and 'extended' social networks rn relation to health and illness.

ScciaLScicnceannddicinc. 12 405-411

Hammer, M., Makiesky-Barrow, S., & Gutwirth, L. (1978). Social networks and
schizophrenia. Schichhrcniafiullctin, A, 522-545.

117

Hansen, W. B. ,,Tobler N. S, & Graham, J. W. (1990) Attrition in substance abuse
prevention research: A meta-analysis of 85 longitudinally followed cohorts.

Exaluaticchxicw 1.4., 677- 685

Hawkins, D. J., Catalano, R. F, & Y, M. J. (1992) Risk and protective factors for
alcohol and other drug problems 1n adolescence and early adulthood: Implications for

substance abuse prevention BsychologicaLBnllctin, 112, 64-105.

Hawkins, J. D., & Catalano, R. F. (1985). Aftercare in drug abuse treatment. International
lcumaLcfithcAddicticns 20. 917-945.

Hawkins, J. D., & Fraser, M. W (1987). The social networks of drug abusers before and
after treatment IntcmaticnallcumalcflthcAddicticns. 22, 343-355.

Higgins, S. T. ,Budney, A. J., Bickel, W. K., & Badger, G. J. (1994). Participation of
significant others in outpatient behavioral treatment predicts greater cocaine

abstinence AmcncanluumalfifflmmdflthLAbusc, 29(1), 47- 56

Hoffman, G. W, Jr., DiRito, D. C. & McGill, E. C. (1993) Three-month follow-up of 28

dual diagnosis inpatients AmcncanlcumMDmundAccthAhusc 12(1),
79-88.

Holden, M. G, Brown, S. A., & Mott, M A. (1988). Social support network of

adolescents: Relation to family alcohol abuse. Amaricanlournalofflnrgand
AlmthAlzusc, 15, 487-498

Holmes-Eber, P. ,& Riger, S. (1990). Hospitalization and the composition of mental
patients' social networks. Schizonhmniafiullctin. 16,157-164.

Hunt, W. A., Barnett, L W. ,& Branch, L. G. (1971) Relapse rates 1n addiction
programs. .lcumachQlinicaLEsrtchclcgy 21 455-456

Institute for Health Policy (1993) Substanccahuscjrcnaticnsnumbcmnchcalth
WWW. Waltham, MA: Brandeis University.

Israel, B. A (1983). Social networks and health status: Linking theory, research, and
practice. Eatirrnscunscllinaandhcalthcducaticn 51(2), 65- 79.

Joe, G. W. ,& Simpson, D. D. (1983). Social factors related to the follow-up status of
opoid addicts. WW3,15, 207-217.

Josephson, E., & Rosen, M. A. (1978). Panel use in a high school drug study. , 115-133.

Kandel, D. B. (1978). Convergences 1n prospective longitudinal surveys of drug use in

normal populations. InD. B. Kandel (Ed), LongitudinachscarchonAmgnsc;

118
EmpiricaLfindingsandmcthcdclcgicalissncs(PP- 3-3 3)- New York: John Wiley &

Sons

Kay, G. R. ,Kalathara, M., & Meiner, A. E. (1989). Diagnostic and behavioral
characteristics of psychiatric patients who abuse substances. HospitaLand

CcmmunitLRsxchiatnr. 90, 1062- 1064

Kessler, R. C. & Price, R. H. (1993). Primary prevention of secondary disorders: A
proposal and agenda Americanlcumalcfficmmuniutfisxchclch. 21, 607-633.

Kleiner, R. J, & Drews, D (1992). Community-based treatment of psychiatric disorders
in the U. S. A. and Norway. Insights for new service delivery systems. International

.IcumalcflScciaLRsychiatry, 38(2), 95- 106

Kosten, T. R., Rounsaville, B. J., & Kleber, H. B. (1983). Concurrent validity of the
addiction severity index IhelcumachNcmcusanndntaLIllncss 111. 606-610.

Laumann, E. 0., & Schumm, L. P. (March 1992). Mcasuringsociaantworkusing
WW? Paper presented at the

Frontiers of social measurement, Chicago, IL.

Lex, B. W. (1991). Some gender differences in alcohol and polysubstance users. Hcalth
Esxchclch. 1.0.121-132.

Lipton, F. R. ,Cohen, C. 1., Fischer, E, & Katz, S. E. (1981). Schizophrenia: A network
crisis Schizophreniafiullctinl 144-151.

Longabaugh, R., & Lewis, D. C. (1988). Key issues in treatment outcome studies.
AlthLflcalthflmcamhflQdd. 12. 163-175.

Lyons, J. E., & McGovern, M. P. (1989). Use of mental health services by dually
diagnosed patients.HQsmtaLand.£Qmmuan£sxch1m,_40, 1067-1069

Macdonald, J. G. (1987). Predictors of treatment outcome for alcoholic women.
IntcmaticnaLlcumalcfltheAddicticns, 22, 235-248

Marlatt, G. A. (1985a). Cognitive factors in the relapse process. In G. A. Marlatt & J. R.
Gordon (Ed), RalaoscPrcxcntioMpp. 128-200). New York, NY: Guilford.

Marlatt, G. A. (1985b). Relapse prevention: Theoretical rationale and overview of the
model. In G. A. Marlatt & J. R. Gordon (Ed.), WWO 3-70). New
York, NY: Guilford.

119

Marlatt. G. A. (1985c). Situational determinants of relapse and skills-training
interventions. In G. A. Marlatt & J. R. Gordon (Ed.), Relapsofimontiompp.
71-127). New York, NY: Guilford.

Marlatt, G A., Baer, J. S. Donovan, D. M, & Kivlahan, D. R. (1988). Addictive
behaviors: Etiology and treatment. WM, 32, 223- 252.

Marlatt. G A., &Gordon.J R (1985) Relapsearexeatioaflaiatmaacestrategieslalk
tneatmeaLofladdiclixeAisatdets New York NY: Guilford

Marsden, P. V. (1990). Network data and measurement. Annualficflmflfiminlm.
1o, 43 5-63.

McCallister, L., & Fischer, C. S. (1983). A procedure for surveying personal networks. In

R S Burt&M J Min0r(Ed) Aaaliedmtmrlaanalysmmkdalogical
inttoduotion(pp. 75- 88). Beverly Hills, CA: Sage.

McGinnis J. M., & Foege, W. H. (1993) Actual causes of deaths 1n the United States.
laumaLofithaAmeacaaMediaaLAssaciatiaa. 229. 2207- 2212.

McLellan, A. T. ,,Luborsky L. ,Woody, G. E., & Druley, K. A. (1983). Predicting
response to drug and alcohol treatments. Role of psychiatric severity. Archmesof

GenetaLEsxchiatm. 40. 620-625

McLellan, A. T., Luborsky, L., Woody, G. E., & O'Brien, C. P. (1980). An improved
diagnostic evaluation instrument for substance abuse patients: The Addiction

Severity Index. IhalaumaloflNemauuadmataLDisease. .168. 26-33.

Mermelstein, R. Lichtenstein, E, & McIntyre, K (1983). Partner support and relapse 1n

smoking-cessation programs. WWW. 5.1..
465-466.

Minkofi‘, K ( 1989). An integrated treatment model for dual diagnosis of psychosis and
addiction HosaitaLaadLammuaithsxchiatm, 30 1031- 1036

Mitchell J. C. (1969). The concept and use of social networks. In J. C. Mitchell (Ed.,)
0.1‘o,101 .01'... 0001'.mottoi n:

Affinanltmms. Manchester, England. Manchester University Press.

Mitchell R. E., & Trickett, E. J. (1980). Task force report: Social networks as mediators
of social support: An analysis of the effects and determinants of social networks.

CommnmtxMeataLHealtUoumal 1.60.). 27-44

Moos, R., & Bliss, F. (1978). Difficulty of follow-up and outcome of alcoholism
treatment laumaLafiStudiesoaAlaokl. 3.2. 473-490

120

Moos, R. H. ,Finney, J. W, & Chan, D. A. (1981). The process of recovery from
alcoholism: 1. Comparing alcoholics patients and matched community controls.

leumaloflfiludiemaAlcokl. 42. 383-402.

Moxley, D. P. (1988) Measuring the social support networks of persons with psychiatric

disabilities: A pilot investigation. Psychosocialfimabilitationloumal, 11(3), 19-27.

Mueller, D. P. (1980). Social networks: A promising direction for research on the
relationship of the social environment to psychiatric disorder. SociaLScicncoand
Medieine. 15A. 147-161.

Narcotics Anonymous (1988). NarcoticoAnonymous. (5th ed.). Van Nuys, CA: World
Service Office.

National Institute on Drug Abuse (1991). Dual diagnosis: Drug abuse and psychiatric

illness, DmgAbusoandflnioAbtisoRcseatch (DHHS Report ADM 91-1704).
Rockville, MI): U. S. Government Printing Office.

Newcomb, M. D., & Bentler, P. M. (1988). Impact of adolescent drug use and social
support on problems of young adults: A longitudinal study. loumaloflAbnonnal
28302110198)!- 21. 64-75.

Nguyen, T. D., Atikisson, C. C., & Stegner, B. L. (1984). Assessment of patient
satisfaction. Exaluatioaoflfimgramfilanaiag. 6. 299-314.

Ohannessian, C. M., & Hesselbrock, V. M. (1993). The influence of perceived social
support on the relationship between family history of alcoholism and drinking
behaviors. Addiction, 88, 1651-1658.

Osher, F. C., & Kofoed, L. L. (1989). Treatment of patients with psychiatric and

psychoactive substance abuse disorders. WM, 510,
1025- 1030.

Oyabu, N, & Garland, T. N. (1987). An investigation of the impact of social support on
the outcome of an alcoholism treatment program. IntcmationaLloumaloflhc
Addictions. 22 221 -.234

Perkins, K A. ,Simpson, J. C. & Tsuang, M. T. (1986). Ten-year follow-up of drug

abusers with acute or chronic psychosis. Hosmtaljndfiommunttlfisxchiatty, 31.
481-484

Phillips, S. L. (1981). Network characteristics related to the well-being of normals: A

comparative base. Schizonhtcniaflullctin, 1, 117-123.

121

Polich, J. M., Armor, D. J ., & Braiker, H. B. (1980). Ihoooutsooflalcoholismgfiout
maficntroatmcntl No. R-2433-NTAAA). Santa Monica, CA: Rand.

Powell, B. J. Penick, E. C., Nickel, E. J., Liskow, B. I. ,Riesenmy, K. D. ,Campion, S.
L., & Brown, E. F. (1992). Outcomes of co-morbid alcoholic men. A 1-year

follow-Up AlcakhsmfimeaLaadExpemneamlResearenm. 131-138

Regier, D. A., Farmer, M. E., & Rae, D. S. (1990). Comorbidity of mental disorders with

alcohol and other drug abuse. laumalofitheAmerieaaMedieaLAssoeiatian. 265.
251 1-2518.

Rhoads, D. L. (1983). A longitudinal study of life stress and social support among drug
abusers. IkJatematiaaaLlaumaLaflheAddictianslfi. 195-222

Ries, R. K. & Ellingson, T. (1990) A pilot assessment at one month of 17 dual diagnosis
patients Hosaitdaadfiammuaitefisychiatm. 41. 1230- 1233

Ritter, C. (1988). Social supports social networks, and health behaviors. In D. S.

Gochman(Ed). me 149-161) New
York, NY: Plenum.

Robins, L., & Smith, E. M. (1980). Longitudinal studies of alcohol and drug problems:

Sex differences. In 0. J. Kalant (Ed), AlcohoLandonigomblcmsjnmmcn (Vol. 5).
New York, NY: Plenum Press.

Rosenberg, H. (1983). Relapsed versus non-relapsed alcohol abusers: Coping skills, life
events, and social support. AddictmoBchayicLs, 8, 183-186.

Ross, H. E., Glaser, F. B., & Germanson, T. (1988). The prevalence of psychiatric
disorders in patients with alcohol and other drug problems. Archixcsofficnctal

28mm. 5.5. 1023-1031.

Rounsaville, B. J., Dolinsky, Z. S., Babor, T. F., & Meyer, R. E. (1987).
Psychopathology as a predictor of treatment outcome in alcoholics. American

leumamflisxehim. 6.4. 505-513.

Rumptz, M H., Sullivan, C. M., Davidson, W. S, 11, & Basta, J. (1991). Anecological
approach to tracking battered women over time. Miolcncoandlictims.

Rydell- C- R. & Everingham. 5- S- (1994).Caatralh1m_eaca1ae_8uaa1ylersus_demaad
programs. Santa Monica, CA: Rand.

Schoenfeld, P., Halevy, J. Hemley-van der Velden, E, & Ruhf, L. (1986). Long-term
outcome of network therapy. HospitaLandLommiinimBsxchiatm, 31, 373-376

122

Selzer, M. L. (1971). The Michigan Alcoholism Screening Test: The quest for a new
diagnostic instrument. AmedmleumaLofiBsyohiatm, 121, 1653-1658.

Shattuck, A. L. White, E. & Kristal, A. R. (1992). How women's adopted low-fat diets
affect their husbands. Amerioaaloumaloiflublioflealth. 82. 1244-1250

Siddall, J. W., & Conway, G. L. (1988). Interactional variables associated with retention

and success in residential drug treatment. IntmationafloumaloflthoAddictions, 23,
1241- 1254.

Simpson, D. D. ,& Sells, S. B. (1983) Effectiveness of treatment for drug abuse: An
overview of the DARP research program. AdyaneeslnAleoholnnifiubstanee
Abuse. 2(1) 7- 29

Smith, E. M. (1993). Race or racism? Addiction in the United States. Annaleof
Epidemiology. 3(2), 165-170.

Smith, J. W ,.Fraw1ey,J P., & Howard, M. O. (1991). Environmental risks to be
considered 1n substance user treatment planning, implementation, and assessment.

IatematioaaLlouraalofltheAddietioas. 26, 371 -375

Sobell, L. C. Sobell, M. B., Toneatto, T., &Leo, G. I. (1993). What tnggers the
resolution of alcohol problems without treatment? Alcoholismflnicalond

EmpeamentaLReseamh. 11, 217-224

Sokolovsky, J ., Cohen, C., Berger, D., & Geiger, J. (1978). Personal networks of
ex-mental patients in a Manhattan SRO hotel, HumanflmanizntioMVol. 37, pp.
5-15).

Sorensen, J. & Gibson, D (1983). Community network approach to drug abuse

treatment BulleuaofltheSoeietyoERsyehologistsJaAddietiyeBehmdom. 2. 99-102

Stephens, R., & Cottrell, E. (1972). A follow-up study of 200 narcotic addicts committed
for treatment under the Narcotic Addict Rehabilitation Act (NARA). Bijtishloumal
ofiAddietion. 61. 45-53.

Strug, D. L., & Hyman, M. M. (1981). Social networks of alcoholics. JournaLQLSttidics
oaAloohol. .42. 855-885.

Tausig, M. (1992). Caregiver network structure, support and caregiver distress. American

lpumaLoLCommuaityJZsychology,20.81-.96

Thoits, P. (1982). Conceptual, methodological and theoretical problems 1n studying

social support as a buffer against life stress. JournalofifiealtnandSocinLBchainor,
23, 145- 159

123

Tuchfeld, B. S. (1981). Spontaneous remission in alcoholics: Empirical observations and

theoretical implications. loumaLoLStudicsonflcohol, 62, 626-641.

Van der Poel, M. G. M. (1993). Delineating personal support networks. Socialflemotks,
15, 49-70.

Wallace, B. C. (1989). Psychological and environmental determinants of relapse in crack

cocaine smokers. loumanLSubstaaoeAbuseIreatmem. 6. 95-106.

Wasylenki, D. A. ,Goering, P. N. Lemire, D. ,Lindsey, S., & Lancee, W. (1993). The
hostel outreach program: Assertive case management for homeless mentally ill

persons. HospitalaadLommuaityEsyehiatm. 4.4.. 848- 853

Westermeyer, J. (1989) Nontreatment factors affecting treatment outcome in substance

abuse. AmeneaaloumalofiflmgaadfleohoLAbuse. 16(1). 13-29

Westermeyer, J. & Neider, J. (1988). Social networks and psychopathology among
substance abusers. AmcncanloumaLoflfisychiatry, 136, 1265- 1269.

Ziedonis, D. M., Rayford, B. 8., Bryant, K. J ., & Rounsaville, B. J. (1994). Psychiatric
comorbidity in white and African-American cocaine addicts seeking substance abuse

treatment. HospitaLaadLommuaityRsyehiatnt .46. 43-49.

Zimmerman, R. S., & Connor, C. (1989) Health promotion in context: The effects of
significant others on health behavior change. Henlthfidncationfluactcdy, 16, 57- 75.

Zobeck, T. S., Stinson, F. S., Grant, B. F. ,&Bertoluci, D. (1993). Itendsjn
3W. Rockville, MD:

National Institute of alcohol abuse and alcoholism.