effects of readiness for drug abuse treatment on client retention and assessment of process

Addiction (1998) 93(8), 1177 ± 1190

RESEARCH REPORT

Effects of readiness for drug abuse treatmenton client retention and assessment of process

GEORGE W. JOE, D. DWAYNE SIMPSON & KIRK M. BROOME

Texas Christian University, Texas, USA

Abstract

Aim s. This study examined client motivation as a predictor of retention and therapeutic engagement across

the major types of treatment settings represented in the third national drug abuse treatment outcome study

(DATOS) conducted in the United States. Design. Sequential admissions during 1991 ± 93 to 37 programs

provided representative samples of community-based treatment populations. Based on this naturalistic

non-experimental evaluation design, hierarchical linear model (HLM) analysis for nested data was used to

contro l for systematic variations in retention rates and client attributes among programs within modalities.

Setting. The data were collected from long-term residential (LTR), outpatient methadone (OMT) and

outpatient drug-free (ODF) programs located in 11 large cities. Participants. A total of 2265 clients in 18

LTR, 981 clients in 13 OM T and 1791 clients in 16 ODF programs were studied . Measurem ents.

Pre-treatment variables included problem recognition and treatment readiness (two stages of motivation),

socio-demographic indicators, drug use history and dependence, criminality, co-morbid psychiatric diagnosis

and previous treatment. Retention and engagement (based on ratings of client and counselor relationships)

served as outcome criteria. Findings. Pre-treatment motivation was related to retention in all three

modalities, and the treatment readiness scale was the strongest predictor in LTR and OMT. Higher treatment

readiness also was signi ® cantly related to early therapeutic engagement in each modality. Conclusions.

Indicators of intrinsic motivation Ð especially readiness for treatment Ð were not only signi ® cant predictors of

engagement and retention, but were more important than socio-demographic, drug use and other background

variables. Improved assessments and planning of interventions that focus on stages of readiness for change and

recovery should help improve treatment systems.

patient methadone (Simpson & Joe, 1993),

therapeutic community (De Leon et al., 1994;

De Leon, Melnick & Kressel, 1997) and tobacco

and alcohol treatments (Cox & Klinger, 1988;

Prochaska, DiClemente & Norcross, 1992;

Ryan, Plant & O’ Malley, 1995). The connec-

tions of motivation to other elements of treat-

ment process also are becoming more clearly

Introduction

The importance of the motivational status of

clients entering drug abuse treatment has been

recognized for many years (Glasscote et al.,

1972), and recent studies have shown this to be

a consistent predictor of client retention.

Speci® cally, low scores on motivational assess-

ments are related to early dropout from out-

Correspondence: George W. Joe, Institute of Behavioral Research, Texas Christian University, TCU Box 298740,Fort Worth, TX 76129 , USA. Tel: 1 1 817 921 7226; Fax: 1 1 817 921 7290 . E-mail: www.ibr.tcu.edu [email protected]

Submitted 14th July 1997 ; initial review completed 25th November 1997; Final version accepted 3rd February1998.

0965 ± 2140 /98/081177 ± 14 $9.50 Ó Society for the Study of Addiction to Alcohol and Other Drugs

Carfax Publishing Limited

1178 George W . Joe et al.

established. For instance, client motivation at

treatment intake is associated with the formation

of better therapeutic relationships (Simpson et

al., 1997f), more favorable perceptions of coun-

selor competence and support from peers

(Broome et al., 1997) and increased session

attendance (Simpson et al., 1997f). Thus, moti-

vational interviewing has been advocated and

used as a tool for strengthening client commit-

ments to recovery in a variety of substance abuse

treatment settings (Miller & Rollnick, 1991).

However, motivation is a complex and

dynamic construct. It includes both extrinsic and

intrinsic dimensions as suggested by the diversity

of reasons given by individuals for entering treat-

ment and quitting drugs. Frequently cited rea-

sons are tiring of the drug-related hustle, ª hitting

bottomº , fear of being jailed and family responsi-

bilities and pressures (Joe, Chastain & Simpson,

1990; Cunningham et al., 1994; Varney et al.,

1995). Although extrinsic motivators such as

legal pressures and sanctions can be useful in

keeping clients in treatment (Pompi & Resnick,

1987; De Leon, 1988; Anglin, Brecht & Mad-

dahian, 1989), intrinsic factors are commonly

considered more fundamental to the recovery

process (Deci & Ryan, 1985; Curry, Wagner &

Grothaus, 1990; Cunningham et al., 1994). Even

among clients legally remanded to treatment, the

extent to which they acknowledge having prob-

lems that are drug related predicts higher levels

of therapeutic engagement (Broome et al., 1997).

Intrinsic factors have received considerable

attention in the last 20 years because of the desire

to understand why some people are successful

while others fail in their attempts to change their

behaviors. Analysis of various reasons given for

wanting to change a behavior and for entering

some type of treatment to implement this has

identi® ed intrinsic factors involved.

Components of motivation also have been

con® gured as progressive stages of cognitive

commitments that help initiate and sustain a

dynamic process of behavioral change. Cashdan

(1973) presents an early conceptual formulation

for such models, but it has only been more

recently that they have received extensive study

when they were adapted to the substance abuse

® eld (Prochaska et al., 1992; De Leon, 1996).

Important phases that are common across these

models include recognition of problems caused

by drug use, an interest and desire for help in

making changes, readiness to enter a formal

process to guide change and action steps that will

help carry out the plan for change (Simpson

& Joe, 1993). Prochaska et al. (1992) describe

the process in ® ve stages (precontemplation,

contemplation, preparation, action and

maintenance), while De Leon (1996) focuses

on six pre-treatment stages and four more

that are related to during-treatment experiences.

De Leon’ s six pre-treatment stages include

denial (no problem recognition or acceptance of

the consequences of continued use), ambivalence

(some problem recognition but inconsistent

acceptance of the consequences of continued

use), extrinsic motivation (attribute their drug-

related problems to external in¯ uences), intrinsic

motivation (inner reasons for personal change),

readiness for change (willingness to seek change

options that are not treatment related), and

treatment readiness (acknowledge treatment as

the only option for change). The four treatment-

related stages are deaddiction, abstinence,

continuance and integration and identity

change.

Because several cognitive readiness stages

appear to precede behavioral change, they have

been given particular attention in several drug

abuse treatment evaluation studies (see Simpson,

1993). According to implications of this type of

model, the success of an individual in treatment

will depend in part upon the stage of change

achieved (Prochaska et al., 1992). More success

would be expected for those who are at later

stages of readiness and commitment. Corre-

spondingly, to minimize the likelihood of early

treatment dropout, it would be helpful to assess

the motivation with which individuals enter

treatment and develop a treatment plan accord-

ingly. The Circumstances, Motivation, Readi-

ness and Suitability (CMRS) components scale

was developed by De Leon & Jainchill (1986)

initially for use in a long-term residential thera-

peutic community environment, and related

work by Simpson & Joe (1993) in outpatient

methadone treatment settings use stage-based

scales representing problem recognition, desire

for help and treatment readiness. Not only is

there evidence that these types of scales can

assess stages of readiness and predict outcomes

for drug abuse treatment populations reliably,

but they also have predictive utility beyond that

offered by other commonly used measures of

client pre-treatment background and function-

ing.

Treatment readiness and retention 1179

Findings from the national multisite Drug

Abuse Treatment Outcome Study (DATOS) in

the United States have replicated results from

earlier evaluations showing that highly signi® cant

behavioral improvements occur from before to

after treatment. Furthermore, these post-

treatment outcomes are related directly to reten-

tion (Hubbard et al., 1997; Simpson, Joe &

Brown, 1997c). The inclusion of motivation

assessment (based on the CMRS) in the exten-

sive client intake battery of this database pro-

vides an excellent opportunity to examine

treatment readiness as a predictor of retention in

three different treatment modalities in this

national sample of treatment programs. That

was the ® rst objective for the present study.

Even more valuable, however, is an examin-

ation of the relationship of the measure of treat-

ment readiness with other indicators of

treatment process. It is the therapeutic process

that is important to recovery. As noted in studies

of drug treatment process components (e.g.

Moos, Finney & Cronkite, 1990; Joe, Simpson &

Hubbard, 1991; Finney, Hahn & Moos, 1996;

Simpson et al., 1997e,f), the treatment environ-

ment, patient needs and delivery of services are

active ingredients in the process that affect out-

come. De Leon (1995) and Simpson (1993)

draw particular attention to the therapeutic rela-

tionships between counselor and client. In pre-

vious research with another data set (Simpson et

al., 1997e), we demonstrated the importance of

treatment engagementÐ de® ned in terms of

counseling session attendance and mutual rat-

ings of the therapeutic relationship between

counselor and clientÐ for retention in a sample

of methadone maintenance clients. Furthermore,

based on previous ® ndings from other studies

(e.g. Ryan et al., 1995; Simpson et al., 1997f),

we would expect to ® nd evidence suggesting that

pre-treatment motivation most likely affects

retention because it improves treatment engage-

ment. This study therefore used a two-step

analytic approach for investigating client

motivational constructs and their relationship

to components in the `black box’ of treat-

ment. First, the generalizability of relationships

between treatment readiness and retention was

examined across programs from three major

modalities. Next, treatment readiness is exam-

ined as a correlate of elements of early thera-

peutic engagement that could help explain

differences in client retention.

Methods

As part of the national Drug Abuse Treatment

Outcome Study (DATOS; see Flynn et al.,

1997) between 1991 and 1993 a total of 10 010

clients were admitted to 96 drug treatment pro-

grams in 11 cities located throughout the United

States. The programs had been sampled pur-

posely in an attempt to assure representative and

naturalistic sources of data. They included 2774

admissions to 21 long-term residential (LTR)

programs, 3122 admissions to 14 short-term

inpatient (STI) programs, 2574 admissions to 32

outpatient drug-free (ODF) programs and 1540

admissions to 29 outpatient methadone treat-

ment (OMT) programs. More detailed method-

ological and general outcome ® ndings from this

evaluation project have been reported elsewhere

(Simpson & Curry, 1997).

Sample

To satisfy our analytical objectives, two restric-

tions were placed on the research sample. First,

the STI modality was not included because

many clients were discharged from residential

care after very brief treatment episodes due to

decisions based on insurance coverage rather

than on client progress and preferences. Sec-

ondly, several small programs (those with

DATOS records on fewer than 50 clients) were

excluded, leaving only programs judged to have

adequate sample representation that would meet

the analytical criteria for this study. With these

restrictions and the requirement that individuals

have complete data on the variables used, the

® nal samples for this study consisted of 2265

clients from 18 LTR programs, 1791 clients

from 16 ODF programs and 981 clients from 13

OMT programs.

Demographic information for the sample is

presented in Table 1. In general, clients were

male (66%), African American (48%) or white

(36%), and had a mean age of 32 years. Cocaine

(including crack) use and alcohol use before

treatment admission were prevalent across all

three modalities, with about half the clients

(54% and 47%, respectively) in the total sample

reporting at least weekly use of these drugs dur-

ing the year prior to admission. Crack, a form of

cocaine smoked in the United States, was used

by 37% on a weekly basis. Its usage differed by

modality, with half (51%) of LTR clients smok-

ing crack compared with about a third (32%) of


Table 1. Description of clients (%)

Modality

LTR ODF OMT Total(N 5 2265) (N 5 1791) (N 5 981) (N 5 5037)

Average age (years) 31 32 37 32Over 30 54 60 84 62Over 35 22 27 55 30

Male 68 67 61 66White 36 31 46 36African American 49 55 34 48Mexican American 13 11 19 13Completed HS or GED 57 60 67 60Married 22 27 41 28Never married 56 49 31 49Separated/divorced/widowed 22 23 28 24Weekly drug use at intake

Alcohol 55 47 29 47Cocaine 66 43 45 54Opioids 20 9 93 30Marijuana 29 26 16 25

Drug dependence at intakeAlcohol 45 37 21 38Cocaine 80 61 41 65

Prior drug treatment 61 48 74 59Criminal justice status 66 58 28 56Criminal justice referral 33 43 2 31

ODF and 16% of OMT clients. The overall

pro® le of drug use also varied across modalities

with weekly opiate use prior to treatment being

almost universal in OMT but rare in other pro-

grams. Similarly, marijuana and alcohol use

before treatment were comparatively more com-

mon for clients in LTR and ODF.

Procedure

Following admission, each client participated in

a two-part intake interview, with sessions occur-

ring approximately 1 week apart. Intake 1

addressed socio-demographic background, edu-

cation, alcohol and drug use history, illegal

involvement and employment. Intake 2 con-

tained assessment modules based on standard

clinical instruments such as the Diagnostic Inter-

view Schedule (DIS; Robins et al., 1981), Com-

posite International Diagnostic Interview (CIDI;

Robins, Wing & Helzer, 1983), and the Symp-

tom Checklist 90 (SCL-90; Derogatis, Lipman

& Covi, 1973); it also contained a modi® ed

version of the CMRS treatment motivation

instrument (De Leon & Jainchill, 1986). Overall,

42% of the clients in the three modalities were

classi® ed as antisocial personality (ASP) and

13% as Axis I depression or anxiety. A higher

percentage of admissions to LTR (51%) were

classi® ed as ASP than in either ODF (35%) or

OMT (34%). Differences in Axis I depression or

anxiety rates were small between LTR (14%),

ODF (12%), and OMT (11%).

Additional interviews were conducted with

each client at the end of the ® rst and third

months of treatment. These in-treatment inter-

views focused on psychological functioning, drug

use, health and employment, as well as speci® c

treatment-related experiences and perceptions.

As described below, these were used in the pre-

sent study to calculate measures of therapeutic

engagement. A subsample of clients was subse-

quently interviewed following treatment, but that

information was not used in the analyses for the

present study.

Variables

Retention criteria. Whether clients stayed at least

90 days was the retention criterion for the LTR

and ODF analyses; 360 days was used as the

criterion for the OMT analyses. These criteria


were de® ned on the basis of previous research

indicating that minimal thresholds of therapeutic

contact are required before outcomes improve

beyond those for intake-only clients (Simpson

1979, 1981; Condelli & Hubbard, 1994). These

thresholds are 3 months for LTR programs

(such as therapeutic communities) and for ODF

programs, but for OMT the period extends up to

a year (Simpson, 1981; Moolchan & Hoffman,

1994; MacGowan et al., 1996). These also

proved to be representative of the median

lengths of retention for clients in DATOS. That

is, half of all LTR clients stayed in treatment at

least 90 days, but with the median stay ranging

from 29 to 177 days for the 18 programs. In

ODF, half of all the clients in this modality also

stayed at least 90 days, and the median stay

ranged from 42 to 144 days for the 16 ODF

programs. For OMT, half of all clients remained

at least 360 days in their methadone treatment,

with the median treatment tenures ranging from

117 to 583 days for the 13 OMT programs. (See

Simpson et al., 1997b, for more information on

program diversity in client retention rates.)

Because of the variation in retention rates within

modality, we used hierarchical linear modeling

(HLM) regression to analyze the relationship

between predictors and retention.

Pre-treatment motivation scales. Three scales,

labeled as problem recognition (PR), desire for

help (DH) and treatment readiness (TR), corre-

sponding to stages of therapeutic readiness

(Simpson & Joe, 1993), emerged from the

psychometric analyses of the 20-item subset

from the CMRS that was included in the

DATOS intake assessment instrument (Flynn et

al., 1997). Problem recognition is viewed as a

measure of the level of personal acknowledge-

ment or denial of behavioral problems that

results from drug use. Desire for help represents

a subsequent step and addresses an awareness of

an intrinsic need for change and a corresponding

interest in getting help. Treatment readiness, a

third stage, addresses degree of commitment to

active change through participation in a treat-

ment program. As examples, the two items with

highest loadings on problem recognition

addressed the fact that drug use was causing

serious problems in the individual’ s life. The two

strongest items on the desire for help scale were

being tired of the ª drug hustleº life and a sense

of urgency to stop drug use. For treatment readi-

ness, the two items most highly related to the

scale acknowledged the importance of treatment

for personal recovery goals.

Answers to each item used a three-point

response scale (ª not at allº , ª agree somewhatº ,

ª agree very muchº ). Separate con® rmatory fac-

tor analyses testing the stages of readiness model

(Simpson & Joe, 1993) and the CMRS compo-

nents model (De Leon & Jainchill, 1986) were

applied to the data in each treatment modality.

Higher ® t indices and reliabilities were found for

the stages of readiness model for the LTR

(GFI 5 0.91 and RMSEA 5 0.070), for the ODF

(GFI 5 0.90; RMSEA 5 0.075) and for the

OMT (GFI 5 0.90; RMSEA 5 0.071) programs.

Each of the three scales had satisfactory

reliability, yielding a coef® cient alpha of 0.71 for

problem recognition (PR), 0.81 for desire for

help (DH), and 0.71 for treatment readiness

(TR). These measures were highly intercorre-

lated (r 5 0.70 between PR and DH; 0.72

between PR and TR, and 0.76 between DH and

TR), but Simpson & Joe (1993) have used struc-

tural equation modeling to establish the linear

sequence of change for these stages. Because the

treatment readiness scale represents a later stage

of cognitive readiness the present study concen-

trates primarily on that scale.

Demographics. Basic demographic variables

were included in the analyses to control for client

differences across treatment programs. Measures

included age at intake, gender, race (being white

vs. non-white), marital status (never married;

separated, divorced or widowed; or married),

and employment in the last 6 months before

admission (full or part time versus none).

Drug use. Drug use history has been shown to

be related to treatment retention and engage-

ment (Hubbard et al., 1989; Joe, Dansereau &

Simpson, 1994; Simpson et al., 1995). As covari-

ates, we therefore included indicators of DSM-

III-R alcohol dependence, at least weekly

marijuana use, and DSM-III-R cocaine depen-

dence. These drugs were targeted because they

represented major problems for clients across all

modalities. We also examined frequency of

cocaine use and opiate use as potential covari-

ates. Cocaine dependence (rather than frequency

of use) was chosen to be included in the ® nal

analyses because of its implications for clinical

severity. Except for methadone treatment, opiate


drug use was reported too infrequently to be

useful as a covariate.

Previous drug treatment. Treatment has been

found to have cumulative effects (Hser et al.,

1997). Therefore, to control for the amount of

treatment exposure we included the total num-

ber of weeks the client reported in all treatment

programs previously attended.

Criminality. Client criminal history and legal

status have rami® cations for treatment retention

(Desmond & Maddux, 1996). For the present

study, we used whether the client had a legal

status at intake (parole, probation, awaiting trial

or sentencing and case pending) and total num-

ber of previous arrests. Information on arrests

was collected using a precoded eight-point scale

(0 5 0, 1 5 1 or 2 times, 2 5 3± 5 times, 3 5 6± 10

times, 4 5 11± 49 times, 5 5 50± 99 times,

6 5 100± 199 times, and 7 5 over 199 times).

Psychiatric diagnosis. Diagnoses of antisocial

personality (ASP), major depression, and gener-

alized anxiety disorder were based on DIS items

and scored using algorithms derived from the

DSM-III-R. These three psychiatric disorders

were combined into a single indicator of comor-

bidity and is referred to as psychiatric diagnosis.

Treatment process measures

At months 1 and 3 during treatment, clients

were asked questions about their con® dence in

treatment progress, rapport and relationship with

the primary counselor, along with participation

in the therapeutic process. These represent some

of the major factors that impact treatment effec-

tiveness. That is, if the client perceives that the

treatment is helpful and chooses to become

engaged, then chances for better post-treatment

outcomes increase (Simpson, Joe & Rowan-Szal,

1997e). Scales for measuring three domains rep-

resenting treatment process were created for

month 1 and month 3.

Con® dence in treatment. Five questions formed

this scale as determined in a factor analysis.

These addressed (1) whether the program

helped, (2) whether the treatment helped stop or

reduce drug use, (3) the degree to which the

treatment helped stop or reduce drug use, (4)

whether the treatment helped with other prob-

lems and (5) the likelihood of completing treat-

ment. The coef® cient alpha reliability for the

scale was 0.68 for raw scores and 0.72 for stan-

dardized scores.

Rapport with counseling process. Five items were

used to construct a measure of therapeutic rap-

port, based on the factor analysis. These

re¯ ected client perceptions about (1) counselor

support of client goals, (2) counselor sincerity,

(3) ability to work together with counselor, (4)

satisfaction with treatment and (5) whether

treatment matched expectations. The coef® cient

alpha reliability for this scale was 0.83 for both

raw and standardized scores.

Therapeutic engagement. The factor analysis

identi® ed ® ve items that represented level of

client engagement in treatment. They indicated

the degree to which the client (1) felt good about

the progress with his/her problems, (2) was

working on his/her problems, (3) was attempting

to change, (4) although not always successful,

was at least doing something about problems

and (5) acceptance of responsibility for prob-

lems. The coef® cient alpha reliability for this

scale was 0.73 for raw scores and 0.74 for stan-

dardized scores.

Hierarchical linear model (HLM ) regression

Clients treated within the same treatment facility

are likely to be similar to one another in many

respects. That is, they usually come from the

same general residential areas and they are

exposed to the same general treatment condi-

tions and philosophy. As a result, statistical com-

parisons that combine data from multiple

programs can yield ® ndings in which client and

program effects are intertwined. A technique

that can take into account correlations among

clients within programs while allowing use of

important individual-level variation is hierarchi-

cal linear model (HLM) regression analysis

(Bryk & Raudenbush, 1992). This procedure

simultaneously estimates equations for each level

of the hierarchical design. In the present study,

for example, clients and programs form the ® rst

and second levels, respectively, meaning that

clients are nested within treatment programs.

Procedurally, for this study, we focused ini-

tially on the hypothesis that client readiness for

treatment predicts program retention rates


within each modality separately (while con-

trolling for retention rate variation among the

programs). We also re-examined this hypothesis

while holding constant certain client characteris-

tics through a ª random coef® cients regressionº

model using HLM (see Bryk & Raudenbush,

1992, p. 20); the client prediction equation

included the intercept, client motivation, a

pro® le of client-level covariates and a random

error term. For the program equation the inter-

cept and slopes for the predictors were allowed

to vary across programs, but no attempt was

made to predict these variations. Finally, the test

of the intercept, as determined from the second

level equation, addressed the following question:

ª If all programs within a modality were treating

the same type of client as delimited by our

covariates, would retention rates differ?º The

logistic model option of the HLM regression was

used in the present study since the dependent

variable was dichotomous. Because of the differ-

ences between modalities in the types of clients

served and the treatment settings involved, all

analyses were restricted to one modality at a

time.

Results

Random coef ® cients HLM regression with treatment

readiness

The ® rst hypothesis tested was that treatment

readiness predicted treatment retention for

at least 90 days in LTR and ODF programs,

and 360 days in OMT. As anticipated, this

was con® rmed in LTR (b 5 0.67, t 5 5.27,

p , 0.0001) while statistically controlling for

between-program variation in retention rates. An

increase of the treatment readiness score by one

on the three-point scale was found to double the

odds (a 96% increase) of staying at least 90 days.

Treatment readiness also proved to be a

signi® cant predictor of 360-day retention in

OMT. The in¯ uence of treatment readiness in

OMT was about the same strength (b 5 0.72,

t 5 2.23, p , 0.045) as in LTR, with a 103%

increase in retention for each unit increase in this

motivation score. In addition, the OMT analyses

suggested that client readiness for treatment was

signi® cantly more important for predicting the

360-day retention criterion in some methadone

programs than in others. Although in the

hypothesized direction, the treatment readiness

scale was not a statistically signi® cant predictor

of 90-day retention in ODF (b 5 0.15, t 5 1.32,

p , 0.20).

Treatment readiness adjusted for covariates

Having established the signi® cance of treatment

readiness as a predictor of the retention criteria

in both LTR and OMT, the next step was to

analyze its signi® cance within the context of

client demographic and background variables. In

LTR, an HLM analysis in which the slopes for

the covariates were the same across programs

(i.e. non-random regression coef® cients) was

supported by the data. The results presented in

Table 2 show that treatment readiness is still a

signi® cant predictor even when adjusted for the

set of covariates, and that a 101% increase in the

90-day retention criterion would result for each

unit increase in the treatment readiness score.

Covariates that were statistically signi® cant

included age, having never been married, being

employed at intake, being alcohol dependent,

using marijuana at least weekly, having a legal

status and number of life-time arrests. Covari-

ates positively associated with retention were age

(a 3.2% increase per year), having never married

(odds ratio of 1.42), being employed (odds ratio

of 1.17), being alcohol dependent (odds ratio of

1.33) and having a legal status (odds ratio of

1.52). Negative in¯ uences on retention included

at least weekly marijuana use (odds ratio of 0.87)

and life-time arrests (12.5% decrease for each

unit increase of the arrest scale).

Similar results for treatment readiness were

observed for OMT treatment, as summarized in

Table 2. Treatment readiness was again

signi® cant, even when adjusted for the pro® le of

covariates. That is, a 110% increase in retention

accompanied each unit increase on the treatment

readiness score. Covariates that increased reten-

tion were age (representing a 2% increase for

each year) and being white (odds ratio of 1.55).

Having a legal status decreased the odds of

retention by 0.75.

In the analysis based on ODF programs, how-

ever, treatment readiness (even when adjusted

for the covariates) was not a signi® cant predictor

of the 90-day retention criterion. We therefore

examined the problem recognition scale (repre-

senting motivation) as a predictor of retention

and found that it did signi® cantly predict reten-

tion (b 5 0.16, SE 5 0.07, t 5 2.19, p , 0.044).

The results showed for each unit of increase on


Ta

ble

2.

Hie

rarc

hic

al

lin

ear

model

regre

ssio

na

na

lysi

sof

trea

tmen

tre

ten

tion

inlo

ng-t

erm

resi

den

tia

la

nd

outp

ati

ent

met

ha

don

etr

eatm

ents

Lo

ng

-ter

mre

sid

enti

al

Ou

tpati

en

tm

eth

ad

on

e

Pre

dic

tor

bS

Et

o/r

%C

han

ge

bS

Et

o/r

%C

han

ge

Inte

rcep

t2

0.3

50

.17

22

.12

20

.06

0.2

22

0.2

5

Tre

atm

ent

read

iness

0.7

00

.15

4.6

6*

**

*1

01

0.7

50

.31

2.4

2*

11

1

Dem

og

rap

hic

sA

ge

0.0

30

.01

6.0

6*

**

*3

.20

.02

0.0

12

.02

*2

.0M

ale

0.0

60

.10

0.6

51

.07

0.0

50

.05

0.9

21

.05

Wh

ite

20

.08

0.1

12

0.6

80

.93

0.4

40

.08

5.2

6**

**

1.5

5N

ever

marr

ied

0.3

50

.09

3.9

7*

**

1.4

22

0.1

70

.17

21

.02

0.8

4S

epara

ted

/div

orc

ed/w

ido

wed

0.0

60

.12

0.5

31

.07

20

.03

0.0

92

0.3

90

.97

Em

plo

yed

0.1

60

.05

2.9

9*

*1

.17

20

.10

0.1

22

0.8

00

.91

Dru

gs

Alc

oh

ol

dep

en

den

ce

0.2

80

.08

3.6

6*

*1

.33

0.0

10

.18

0.0

31

.01

Mari

juan

a(a

tle

ast

week

ly)

20

.14

0.0

72

2.1

0*

0.8

70

.13

0.1

60

.79

1.1

4C

oca

ine

dep

en

den

ce

0.0

90

.10

0.8

41

.09

0.1

50

.11

1.3

51

.16

Pre

vio

us

treatm

en

tA

mo

un

to

fd

rug

treatm

en

t(w

eek

s)0

.00

0.0

00

.72

0.0

0.0

00

.00

1.2

60

.1

Cri

min

ali

tyL

egal

statu

s0

.42

0.1

13

.77

**

1.5

22

0.3

00

.15

22

.05

*0

.74

Arr

ests

20

.13

0.0

32

4.1

9*

**

*2

12

.40

.02

0.0

60

.37

2.1

Psy

chia

tric

dia

gn

osi

s2

0.0

00

.07

20

.04

1.0

02

0.1

40

.13

21

.09

0.8

7

*p

,0

.05

;*

*p

,0

.01

;**

*p

,0

.00

1;

***

*p

,0

.00

01

.


Ta

ble

3.

Corr

ela

tion

sof

trea

tmen

tre

ad

ines

sw

ith

pro

cess

va

ria

ble

sby

mod

ali

tya

nd

rete

nti

on

clu

ster

s

LT

Rre

ten

tio

ncl

ust

ers

Lo

wH

igh

Mo

1M

o3

Mo

1M

o3

( N5

91

1)

( N5

53

7)

( N5

90

7)

( N5

61

1)

Co

n®

den

ce0

.41

**

0.3

4*

*0

.39

**

0.2

8*

*R

ap

po

rt0

.32

**

0.1

7*

*0

.29

**

0.2

2*

*E

ngag

emen

t0

.34

**

0.2

5*

*0

.24

**

0.2

2*

*

OM

Tre

ten

tio

nclu

ster

s

Lo

wM

idd

leH

igh

Mo

1M

o3

Mo

1M

o3

Mo

1M

o3

( N5

93

)( N

57

0)

( N5

45

8)

( N5

38

1)

( N5

24

3)

( N5

22

0)

Co

n®

den

ce0

.29

*0

.36

*0

.18

**

0.1

4*

0.1

6*

0.2

2*

Rap

po

rt0

.00

0.1

70

.15

*0

.16

*0

.05

0.0

8E

ngag

emen

t0

.34

**

0.3

4*

0.2

4**

0.2

3*

*0

.31

**

0.2

7*

*

OD

FR

ete

nti

on

clu

sters

Lo

wL

ow

±M

idd

leH

igh

±m

idd

leH

igh

Mo

1M

o3

Mo

1M

o3

Mo

1M

o3

Mo

1M

o3

( N5

33

8)

( N5

11

2)

( N5

34

6)

( N5

17

7)

( N5

51

9)

( N5

23

8)

( N5

19

0)

( N5

15

2)

Co

n®

den

ce0

.19

**

0.0

80

.47

**

0.3

3**

0.4

4*

*0

.43

**

0.4

0*

*0

.42

**

Rap

po

rt0

.18

*2

0.1

00

.29

**

0.3

0**

0.2

7*

*0

.19

*0

.41

**

0.3

8*

*E

ngag

emen

t0

.24

**

0.0

00

.30

**

0.3

7**

0.3

9*

*0

.41

**

0.3

4*

*0

.37

**

*p

,0

.01

;**

p,

0.0

01

.


the three-point problem recognition scale there

would be a 17% increase in retention. The only

other signi® cant predictor in this ODF analysis

was previous weeks of drug treatment (b 5 0.002,

SE 5 0.0007, t 5 2.67, p , 0.018). That is, there

was a 0.2% increase in retention for each addi-

tional week of prior treatment.

Retention rate variation across programs: random

effects for intercepts

HLM analysis was also used to test the hypoth-

esis that retention rates for programs were still

signi® cantly different after adjustments were

made for differences in the types of clients

they admitted. Technically, this involved testing

the estimated variance component for the

intercept; and it was found to be signi® cant

in LTR (variance component 5 0.21, c 2 (17,

N 5 2265) 5 107.67, p , 0.001), OMT (variance

component 5 0.77, c 2 (12, N 5 981) 5 122.63,

p , 0.001) and ODF (variance compo-

nent 5 0.55, c 2 (15, N 5 1791) 5 157.68,

p , 0.001). This implies that, even when

adjusted for the client-level attributes repre-

sented by demographic and background vari-

ables, differences in the retention rates among

programs cannot be explained by sampling varia-

tions. That is, there are other aspects of these

programs that are involved in determining client

retention rates, presumably including variations

in their clinical effectiveness. This was therefore

the next issue addressed in the present study.

Treatment readiness and during-treatment process

measures

The correlations of treatment readiness with

indicators of treatment process measured at

treatment months 1 and 3 are presented in Table

3. These include client ratings of con® dence in

treatment, rapport with the counseling process,

and therapeutic (i.e., clinical) engagement. From

multiple correlation analyses based on the total

LTR sample, treatment readiness was shown

to be correlated signi® cantly with the three

measures of treatment process at month 1

(R2 5 0.19, F (3, 1910) 5 148.44, p , 0.0001)

and month 3 (R2 5 0.12, F (3, 1199) 5 55.12,

p , 0.0001). Similar results occurred for OMT

at month 1 (R 2 5 0.09, F (3, 1054) 5 35.21,

p , 0.0001) and month 3 (R2 5 .09, F (3,

845) 5 27.70, p , 0.0001). However, the

strongest relationships were in ODF at both

month 1 (R2 5 0.25, F (3,1551) 5 174.86,

p , 0.0001) and month 3 (R2 5 0.23, F (3,

800) 5 77.91, p , 0.0001). In order to address

relationships of treatment readiness with these

process indicators across programs which dif-

fered widely in their ability to retain clients in

treatment, these correlations are presented sepa-

rately for groups of programs clustered in our

previous research according to overall retention

pro® les (Simpson et al., 1997b). An indication of

the variability among the programs to retain

clients is given by the percentages of clients in

each cluster who remained in treatment beyond

the minimally effective length of stay. There

were two clusters of LTR programs (lowÐ 46%;

highÐ 61%), three clusters of OMT programs

(low Ð 21%; middleÐ 55%; highÐ 72%), and

four clusters of ODF program (lowÐ 29%; low-

middleÐ 48%; high-middleÐ 61%; highÐ 74%).

As shown in Table 3, treatment readiness is

positively related to all three measures of process

at months 1 and 3 for both low and high reten-

tion LTR programs. For OMT, treatment readi-

ness was consistently related to ratings of

con® dence in treatment and to therapeutic

engagement in all retention clusters. However,

the rapport measure was related to treatment

readiness only in the middle retention OMT

cluster. In ODF, treatment readiness was posi-

tively related to the three process measures for

months 1 and 3 in the low± middle, high± middle

and high retention clusters. In the low retention

ODF cluster, treatment readiness was related

only to the process measures at month 1.

Discussion

In this study we tested the relationship of motiv-

ation for treatment, as measured primarily by

treatment readiness, to critical lengths of stay in

three major treatment modality settings. These

time periods were 90 days for LTR and ODF

and 360 days for OMT. Previous research estab-

lished these periods of time to be predictive of

favorable post-treatment outcomes. Data for this

study came from multiple treatment sites located

throughout the United States using a naturalistic

and longitudinal design, which therefore adds

considerable generalizability and external validity

to the ® ndings (see Finney et al., 1996). The

statistical methodology employed allowed us to


take into consideration the fact that clients are

ª nestedº within programs in large multisite

evaluation designs.

As expected, we found that treatment readi-

ness is an important predictor of 90-day reten-

tion in LTR and 360 days in OMT; although

this relationship was not signi® cant in ODF, we

did ® nd that another motivation scale (problem

recognition) was. When we adjusted the analyses

for client demographic and background infor-

mation, indicators of pre-treatment motivation

not only remained signi® cant in each treatment

population studied, but it was found to be the

most important predictor. In general, an increase

of one unit on the three-point scale doubled the

odds of a client remaining for the crucial mini-

mal length of stay in the LTR and OMT treat-

ment modalities. This ® nding is important

clinically because motivation for treatment is

amenable to change. There are several interven-

tions targeting cognitive appraisal processes that

have shown promise, including motivational

interviewing (Miller & Rollnick, 1991) as well as

cognitive approaches such as node-link mapping

(e.g. Dansereau, Joe & Simpson, 1993; Joe et al.,

1994; Simpson, Dansereau & Joe, 1997a).

However, our ® ndings for the ODF modality

were not as consistent as for the other two

modalities. A plausible reason for this may be

that ODF programs were highly diverse and their

records on treatment services and participation

were poorest in quality of any of the modalities

represented. This added complications to the

calculation of number of days clients remained

in treatment (Simpson et al., 1997c). Another

reason may be that many of the ODF clients

were not ready for treatment but were at an

earlier stage in the stages-of-change model. This

is supported by our ® ndings that problem recog-

nition was a predictor even though the treatment

readiness scale was not. It is reasonable to expect

other factors not measured in this study were

involved in this complex process of change.

We posited that readiness for treatment can

affect treatment engagement. Support for this

hypothesis emerged when we correlated treat-

ment readiness scores with client ratings of treat-

ment con® dence, counseling rapport and

therapeutic engagement. Indeed, positive rela-

tionships were generally found to occur between

our measure of pre-treatment motivation and

these client-rated measures of treatment process

in all three modalities. This ® nding is of particu-

lar importance in ODF because we were able to

establish only a very early stage motivation-to-

retention link for that treatment in this study. It

is the ª process of treatmentº that should affect

post-treatment outcomes, and the motivation

effects on process were supported even for ODF.

We also found the link between treatment readi-

ness and engagement to hold for clients who stay

in treatment, regardless of the overall program

retention rate. That is, the results were the same

in both low and high retention programs. An

implication is that high client motivation can

compensate for less ef® cient treatment to some

extent, but also that less motivated clients may

need early retention initiatives such as motiva-

tional interviewing, node-link mapping or role

induction interviews. These ® ndings demon-

strate the importance of treatment programs

continuously monitoring the reactions of their

clients to the treatment.

The results concerning motivation are consist-

ent with previous research showing it is a predic-

tor of lengths of treatment stay that promote

posttreatment improvement. In particular, it

supports ® ndings that low pre-treatment motiv-

ation is a predictor of early dropout from metha-

done treatment (Simpson & Joe, 1993), from

therapeutic communities (De Leon et al., 1994)

and from tobacco and alcohol treatments (Cox &

Klinger, 1988; Prochaska et al., 1992). This

study adds more evidence for a general relation-

ship found between increased pre-treatment

motivation and a better therapeutic relationship

between client and counselor across modalities

(Dansereau et al., 1993; Simpson et al., 1997a,f).

In addition, it points to the importance of study-

ing indicators for therapeutic alliance (e.g.

Horvath & Luborsky, 1993) because much of the

effectiveness of treatment is believed to revolve

around the development of a strong therapeutic

relationship.

As we have shown in a series of studies from

another research project (Simpson et al., 1997d),

pre-treatment motivation is a prominent factor

in predicting post-treatment outcomes, operating

through its effects on treatment process vari-

ables. Speci® cally, structural equation modeling

indicated that pretreatment motivation was

related to session attendance and the therapeutic

relationship between counselor and client, either

directly or indirectly. This held for predicting

during-treatment drug use, retention (Simpson

et al., 1997f) and post-treatment outcomes


(Simpson et al., 1997e). It is reasonable to

expect the same to be true for clients being

studied in DATOS.

Previous research suggests that intrinsic moti-

vators are better predictors of improved behav-

ioral outcomes than extrinsic motivators (Deci &

Ryan, 1985; Curry et al., 1990; Cunningham et

al., 1994; Ryan et al., 1995). Although the analy-

sis of post-treatment outcomes was beyond the

immediate scope of this study, we did ® nd that

treatment readiness was the strongest predictor

of retention from a set of intake variables that

included legal status, arrest history, amount of

previous drug treatment and employment

statusÐ all extrinsic motivators.

Our present study, as well as other research in

the DATOS, is a prelude to modeling these data

in terms of linear structural equations. This

study and the others are needed for understand-

ing the relationships among the variables and for

later positing how the client background, treat-

ment process, retention and outcomes are inter-

related. These studies also will allow us to

examine the in¯ uence of treatment variability on

these relationships and to develop strategies for

dealing with some of the possible confounds to

interpretation.

In conclusion, this study yields evidence for

the role of treatment readiness in retaining

clients in treatment. In particular, it supported

the generalizability of this relationship under

varying conditions, including modality and pro-

grams. Previously, the evidence for this hypoth-

esis was based on much more limited conditions

with respect to types of programs and client

populations. These ® ndings point to the import-

ance of accurately assessing the stage-of-change

of the client upon treatment intake and to the

possible use of motivation-enhancing induction

procedures when needed. This should improve

the chances of clients becoming engaged in treat-

ment and subsequently enhance their prospects

for rehabilitation. As a consequence, this will

also lead to the treatment program making better

uses of its resources, both human and monetary.

Acknowledgements

This work was supported by National Institute on

Drug Abuse (NIDA) Grant U01-DA10374 as

part of a Cooperative Agreement on the Drug

Abuse Treatment Outcome Study (DATOS).

The project includes a Coordinating DATOS

Research Center (Robert L. Hubbard, Principal

Investigator at NDRI) and two Collaborating

DATOS Research Centers (M. Douglas Anglin,

Principal Investigator at UCLA, and D. Dwayne

Simpson, Principal Investigator at TCU) to con-

duct treatment evaluation studies in association

with NIDA (Bennett W. Fletcher, Principal

Investigator at NIDA). The interpretations and

conclusions contained in this paper do not necess-

arily represent the position of other DATOS

Research Centers, NIDA, or the Department of

Health and Human Services.

ReferencesANGLIN, M. D., BRECHT , M. L. & M ADDAHIAN , E.

(1989) Pretreatment characteristics and treatmentperformance of legally coerced versus voluntarymethadone maintenance admissions, Criminology,27, 537± 557.

BRYK , A. & RAUDENBUSH, S. (1992) Hierarchical Linear

Models: applications and data analysis methods (New-bury Park, CA, Sage Publications).

BRO OME, K. M., KNIGHT, D. K., KNIGHT, K., HILLER,M. L. & SIM PSON , D. D. (1997) Peer, family, andmotivational in¯ uences on drug treatment processand recidivism for probationers, Journal of Clinical

Psychology, 53, 387± 397.CASH DAN , S. (1973) Interactional Psychotherapy: stages

and strategies in behavioral change. (New York, Grune& Stratton).

CONDELLI, W. S. & HUBBARD , R. L. (1994) Relation-ship between time spent in treatment and clientoutcomes from therapeutic communities, Journal of

Substance Abuse Treatment, 11, 25± 33.COX, M. & KLING ER, E. (1988) Motivational model of

alcohol use, Journal of Abnormal Psychology, 97, 168±180.

CUNNINGH AM , J. A., SOBELL, L. C., SOBELL, M. B. &GASKIN , J. (1994 ) Alcohol and drug abusers’ reasonsfor seeking treatment, Addictive Behaviors, 19, 691±696.

CURRY, S. J., WAGNER , E. H. & GROTHAU S, L. C.(1990) Intrinsic and extrinsic motivation for smok-ing cessation, Journal of Consulting and Clinical Psy-

chology, 58, 310± 316.DANSERE AU , D. F., JOE, G.W. & SIMPSON , D. D.

(1993) Node-link mapping: a visual representationstrategy for enhancing drug abuse counseling, Jour-

nal of Counseling Psychology, 40, 385 ± 395.DECI, E. L. & RYAN , R. M. (1985) Intrinsic Motivation

and Self-determination in Human Behavior (NewYork, Plenum).

DE LEON , G. (1988) Legal pressure in therapeuticcommunities, in: LEUKFELD , C. G. & TIMS, F. M.(Eds) Compulsory treatment of drug abuse: research and

clinical practice, DHHS publication no. ADM 88±1578, pp. 160± 177 (Rockville, MD, National Insti-tute on Drug Abuse).

DE LEO N, G. (1995) Therapeutic communities foraddictions: a theoretical framework, International

Journal of the Addictions, 30, 1603 ± 1645.


DE LEON , G. (1996) Integrative recovery: a stageparadigm, Substance Abuse, 17, 51± 63.

DE LEON , G. & JAINCHILL, N. (1986 ) Circumstance,motivation, readiness, and suitability as correlates oftreatment tenure, Journal of Psychoactive Drugs, 18,203± 208.

DE LEON , G., M ELNICK , G. & KRESSEL, D. (1997)Motivation and readiness for therapeutic communitytreatment among cocaine and other drug abusers,American Journal of Drug and Alcohol Abuse, 23,169± 189.

DE LEO N, G., MELNIC K, G., KRESSEL, D. & JAINC HILL,N. (1994) Circumstances, motivation, readiness,and suitability (the CMRS scales): predicting reten-tion in therapeutic community treatment, American

Journal of Drug and Alcohol Abuse, 20, 495 ± 515.DEROG ATIS, L. R., L IPMAN , R. S. & COVI, L. (1973)

The SCL-90: an outpatient psychiatric rating scale,

Psychopharmacology Bulletin, 9, 13± 28.DESMOND , D. P. & MADDU X, J. F. (1996) Compulsory

supervision and methadone maintenance, Journal of

Substance Abuse Treatment, 13, 79± 83.FINNEY , J. W., HAHN , C. & MO OS, R. H. (1996) The

effectiveness of inpatient and outpatient treatmentfor alcohol abuse: the need to focus on mediatorsand moderators of setting effects, Addiction, 91,1773± 1796.

FLYNN , P. M., CRADDOC K, S. G., HUBBARD , R. L.,ANDERSO N, J. & ETHERID GE, R. M. (1997 ) Method-ological overview and research design for theDrug Abuse Treatment Outcome Study (DATOS),

Psychology of Addictive Behaviors, 11, 230 ± 243.GLASSCO TE, R., SUSSEX , J. N., JAFFE, J. H., BALL, J. &

BRILL, L. (1972) The Treatment of Drug Abuse: pro-

grams, problems, prospects (W ashington, DC, Ameri-can Psychiatric Association Division of PublicAffairs, The Joint Information Service).

HO RVATH , A. O. & LUBORSKY , L. (1993 ) The role ofthe therapeutic alliance in psychotherapy, Journal of

Consulting and Clinical Psychology, 61, 561 ± 573.HUBBARD , R. L., MARSDEN , M. E., RACHAL, J. V., HAR-

WOO D, H. J., CAVANAU GH , E. R. & G INZBURG, H. M.(1989) Drug Abuse Treatment: a national study of

effectiveness (Chapel Hill, NC, University of NorthCarolina Press).

HUBBARD , R. L., CRADDOC K, S. G., FLYNN , P. M.,ANDERSO N, J. & ETHERID GE, R. M. (1997) Overviewof 1-year follow-up outcomes in the Drug AbuseTreatment Outcome Study (DATOS), Psychology of

Addictive Behaviors, 11, 261± 278.HSER, Y., ANG LIN, M. D., GRELLA , C. E., LONG SHORE ,

D. & PRENDER GAST, M. L. (1997) Drug treatmentcareers: a conceptual framework and existingresearch ® ndings, Journal of Substance Abuse Treat-

ment, 14, 543± 558.JOE, G. W., CH ASTAIN , R. L. & SIMPSON , D. D. (1990)

Reasons for addiction stages, in: SIMPSO N, D. D. &SELLS, S. B. (Eds) Opioid Addiction and Treatment: a

12-year follow-up, pp. 73 ± 102 (Malabar, FL, KriegerPublishing).

JOE, G. W., DANSERE AU , D. F. & SIMPSON , D. D.(1994) Node-link mapping for counseling cocaineusers in methadone treatment, Journal of Substance

Abuse, 6, 393± 406.

JOE, G. W., SIMPSON , D. D. & HUBBARD , R. L. (1991 )Treatment predictors of tenure in methadonemaintenance, Journal of Substance Abuse, 3, 73± 84.

MACGOW AN , R. J., SWANSON , N. M., BRACKBILL, R.M., RUGG, D. L., BARKER , T. & M OLDE , S. (1996)Retention in methadone maintenance treatment pro-grams, Connecticut and Massachusetts, 1990 ± 1993,Journal of Psychoactive Drugs, 28, 259± 265.

M ILLER, W. R. & ROLLNICK, S. (1991) Motivational

Interviewing: preparing people to change addictive

behavior (New York, Guilford Press).MOO LCHAN , E. T. & HOFFM AN , J. A. (1994 ) Phases of

treatment: a practical approach to methadonemaintenance treatment, International Journal of the

Addictions, 29, 135± 160.MOO S, R. H., FINNEY, J. W. & CRONKIT E, R. C. (1990 )

Alcoholism Treatment: context, process and outcome(New York, Oxford University Press).

POMPI, K. F. & RESNIC K, J. (1987 ) Retention of court-referred adolescents and young adults in the thera-peutic community, American Journal of Drug and

Alcohol Abuse, 13, 309± 325.PROCH ASKA , J. O., D ICLEM ENTE , C. C. & NORCROSS,

J. C. (1992) In search of how people change: appli-cations to addictive behaviors, American Psychologist,47, 1102 ± 1114.

ROBINS, L. N., HELZER , J. E., CROUGHAN , J. & RAT-

CLIFF, K. S. (1981) National institute of mentalhealth diagnostic interview schedule, Archives of

General Psychiatry, 38, 381 ± 389.ROBINS, L. N., W ING , J. K. & HELZER , J. E. (1983 )

Composite International Diagnostic Interview (CIDI)

(Geneva, Switzerland, World Health Organization).RYAN , R. M., PLANT , R. W. & O ’MALLEY , S. O.

(1995) Initial motivations for alcohol treatment:relations with patient characteristics, treatmentinvolvement, and dropout, Addictive Behaviors, 20,279± 297.

SIMPSO N, D. D. (1979) The relation of time spent indrug abuse treatment to posttreatment outcome,

American Journal of Psychiatry, 136, 1449 ± 1453 .SIMPSO N, D. D. (1981) Treatment for drug abuse:

follow-up outcomes and length of time spent,

Archives of General Psychiatry, 38, 875 ± 880.SIMPSO N, D. D. (1993) Drug treatment evaluation

research in the United States, Psychology of Addictive

Behaviors, 7, 120 ± 128.SIMPSO N, D. D. & CURRY, S. J. (Eds) (1997 ) Special

issue: Drug Abuse Treatment Outcome Study, Psy-

chology of Addictive Behaviors, 11.SIMPSO N, D. D. & JO E, G. W. (1993) Motivation as a

predictor of early dropout from drug abuse treat-ment, Psychotherapy, 30, 357 ± 368. [On-line datainstruments]. Available: www.ibr.tcu.edu

SIMPSO N, D. D., JO E, G. W., ROW AN-SZAL, G. A. &GREENER , J. M. (1995) Client engagement andchange during drug abuse treatment, Journal of Sub-

stance Abuse, 7, 117± 134.SIMPSO N, D. D., DANSERE AU , D. F. & JOE, G. W.

(1997a) The DATAR project: cognitive andbehavioral enhancements to community-basedtreatments, in: T IMS, F. M., INC IARDI, J. A.,FLETC HER, B. W. & HORTON , A. M. Jr (Eds) The

Effectiveness of Innovative Approaches in the Treatment


of Drug Abuse, pp. 182 ± 203 (Westport, CT, Green-wood Press).

SIMPSON , D. D., JOE, G. W., BROOM E, K. M., HILLER,M. L., KNIGHT, K. & ROW AN-SZAL, G. A. (1997b)Program diversity and treatment retention rates inthe Drug Abuse Treatment Outcome Study(DATOS), Psychology of Addictive Behaviors, 11,279± 293.

SIMPSON , D. D., JO E, G. W. & BROW N, B. S. (1997c)Treatment retention and follow-up outcomes in theDrug Abuse Treatment Outcome Study (DATOS),Psychology of Addictive Behaviors, 11, 294 ± 307.

SIMPSON , D. D., JOE, G. W., DANSERE AU , D. F. &CH ATH AM , L. R. (1997d) Strategies for improvingmethadone treatment process and outcomes, Journal

of Drug Issues, 27, 239± 260.

SIMPSO N, D. D., JOE, G. W. & ROW AN-SZAL, G. A.(1997e ) Drug abuse treatment retention and processeffects on follow-up outcomes, Drug and Alcohol

Dependence , 47, 227± 235.SIMPSO N, D. D., JO E, G. W., ROW AN-SZAL, G. A. &

GREENER , J. M. (1997f) Drug abuse treatment pro-cess components that improve retention, Journal of

Substance Abuse Treatment, 14, 565 ± 572.VARNEY , S. M., ROHSENO W , D. J., DEY, A. N.,

MYERS, M. G., ZWICK, W. R.& M ONTI, P. M.(1995) Factors associated with help seeking andperceived dependence among cocaine users,

American Journal of Drug and Alcohol Abuse, 21,81± 91.

effects of readiness for drug abuse treatment on client retention and assessment of process

Documents