THE PRISON JOURNAL / September 1999Martin et al. / THERAPEUTIC COMMUNITY TREATMENT

THREE-YEAR OUTCOMES OF THERAPEUTIC COMMUNITY

TREATMENT FOR DRUG-INVOLVED OFFENDERS IN DELAWARE: FROM PRISON

TO WORK RELEASE TO AFTERCARE

STEVEN S. MARTIN CLIFFORD A. BUTZIN CHRISTINE A. SAUM JAMES A. INCIARDI University of Delaware

Delaware researchers have argued for a continuum of primary (in prison), secondary (work release), and tertiary (aftercare) therapeutic community (TC) treatment for drug-involved offenders. Previous work has demonstrated significant reductions in relapse and recidivism for offenders who received primary and secondary TC treat- ment 1 year after leaving work release. However, much of the effect declines signifi- cantly when the time at risk moves to 3 years after release. Further analyses reveal that program effects remain significant when the model takes into account not simply exposure to the TC program, but, more importantly, program participation, program completion, and aftercare. Clients who complete secondary treatment do better than those with no treatment or program dropouts, and those who receive aftercare do even better in remaining drug- and arrest-free.

Meeting the treatment needs of drug-involved offenders in the correc- tional system would appear to be important in reducing the burgeoning costs of incarceration and related criminal justice and social costs. Researchers have documented the high costs of drug abuse, not only to individual drug

This research was supported by NIDA Grant DA06124 and by NIJ Cooperative Agreement #97-RT-VX-K004. We wish to thank Michael Antonio, Tiho Enev, Druretta Nichols-Jennings, and Daniel O’Connell for their help with the data. We also wish to thank the reviewers for their helpful comments. Correspondence concerning this article should be addressed to Steven Mar- tin, Center for Drug and Alcohol Studies, University of Delaware, Newark, DE 19716-2582; e- mail: martin@udel.edu.

THE PRISON JOURNAL, Vol. 79 No. 3, September 1999 294-320 © 1999 Sage Publications, Inc.

294

users, but to taxpayers and communities as a whole (Harwood, Napolitano, Kristianses, & Collins, 1984; Rice, Kelman, Miller, & Dunmeyer, 1990). These costs are especially significant for drug-involved offenders whose poor health status results in the use of expensive public health services, and whose cycles of relapse and recidivism result in repeated contacts with the criminal justice system (Leukefeld, Logan, Martin, Purvis, & Farabee, 1998). Analyses of 26,000 drug users contacted during the National AIDS Demonstration Research Program during 1987 through 1992 (Auerbach, Wypijewska, & Brodie, 1994), for example, found that those who had been incarcerated had significantly higher rates of drug use in general, and multi- ple drug use, daily drug injections, and unsafe needle use in particular (Inci- ardi, McBride, Platt, & Baxter, 1993). These and other data suggest that chronic drug users are found in the greatest concentrations among inmate populations (Chaiken, 1989; Leukefeld & Tims, 1988).

The rationale for improving outcomes is fundamental—any success in reducing drug use leads to reductions in criminal behavior, which, in turn, generates cost savings to other sectors of society. The major impetus for treatment alternatives in prison settings emerged during the mid-1980s in an attempt to stem the growing criminal justice costs, particularly those related to the custody and health care associated with rapidly increasing numbers of incarcerated drug users (Currie, 1998; Inciardi & Martin, 1993). However, until the 1990s, only a few studies had addressed either the clinical- or cost- effectiveness of drug abuse treatment (Apsler & Harding, 1991; Harwood, Hubbard, Collins, & Rachal, 1988; Hubbard et al., 1989). A recent and widely cited study, however, reported yearly savings of more than a billion dollars resulting from treatment, with the largest savings coming from reduc- tions in crime and health care costs (California Department of Drug and Alcohol Programs, 1994).

Cost-effectiveness or cost savings are what concern policy makers most; but all studies of cost-effectiveness are dependent on the availability of out- come data. Apsler (1991) has listed factors that singly or jointly would improve treatment outcome research: measures on the variability among treatment programs, long project periods, objective validation of self-report measures, the cooperation of the treatment programs, large samples, multiple measures of treatment experience, and multiple measures of outcomes. Recently, Knight, Hiller, and Simpson (1999) have reiterated this list and added to it the need for common or at least comparable indicators across studies.

Correctional systems have long since recognized the need for identifying concentrations of drug users, implementing treatment programs to “captive” populations, and assessing drug treatment outcomes (Petersen, 1974).

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 295

However, the systematic assessment of specific, promising, effective, and replicable drug treatment programs for offenders did not advance concomi- tantly with the increases in inmate populations and the number of new pro- grams. The dilemmas of conducting field studies that can be rigorously evaluated proved difficult to solve. Consequently, most assessments of pro- gram effectiveness were process rather than outcome oriented and did not incorporate multiple outcome criteria. Outcome research, when attempted, involved short follow-up time frames and included only limited use of com- parison groups, standardized measurement instruments, multivariate mod- els, and appropriate control variables (De Leon, Inciardi, & Martin, 1995; Fourcier, 1991; Rouse, 1991; Wexler, 1995).

Many of these shortcomings were addressed through research initiated during the early 1990s, when a number of studies began reporting encourag- ing initial results from the use of therapeutic community (TC) treatment pro- grams for drug-involved offenders. Furthermore, the great majority of these TC programs were prison based (Graham & Wexler, 1997; Lipton, 1995). However, a few programs have extended treatment into the community after release from prison—either through transitional or work release programs, aftercare, or a combination of both (Hiller, Knight, & Simpson, in press; Inci- ardi & Lockwood, 1994; Knight, Simpson, & Hiller, 1996; Mello, Pechan- sky, Inciardi, & Surratt, 1997; Nielsen & Scarpitti, 1997; Wexler, De Leon, Thomas, Kressel, & Peters, 1999). Many of the studies reporting the effec- tiveness of prison therapeutic communities in reducing relapse and recidi- vism have documented significant effects, but usually over relatively short terms—6 to 12 months after release from prison (Lipton, 1995).

With respect to longer term study, in an earlier article the authors exam- ined treatment effectiveness in the Delaware correctional system in terms of relapse and recidivism using a quasi-experimental comparison of four client groups. Because only two of the groups had been randomly assigned, a multi- variate analysis was designed to control for group differences. The findings from those studies suggested significant improvements in outcome for those with both in-prison and postrelease TC treatment, at both 6 months (Martin, Butzin, & Inciardi, 1995) and at 18 months (Inciardi, 1996a, 1996b; Inciardi, Martin, Butzin, Hooper, & Harrison, 1997; Mathias, 1995) after release from prison. These results were cited in federal legislation designed to increase the availability of Residential Substance Abuse Treatment for offenders. Attached to the legislation was $275 million for the period 1997 through 2001 (Office of Justice Programs, 1996).

After reviewing the 1-year findings reported previously for treatment out- come for drug-involved offenders in the Delaware correctional system, we

296 THE PRISON JOURNAL / September 1999

proceed to examine longer term outcomes. The first analyses review client outcomes at 3 years after completing work release. The second analyses also examine 3-year outcomes but take into account the extent to which clients completed a continuum of treatment during prison, work release, and correc- tional supervision in the community.

BACKGROUND

Delaware’s corrections-based multistage TC treatment program began in 1991 as a research demonstration project funded by the National Institute on Drug Abuse (NIDA), and is now a continuing program in the Delaware cor- rectional system. Like other therapeutic communities, it was designed to pro- vide a total treatment environment in which a drug user’s transformations in behavior, attitudes, emotions, and values are introduced and inculcated. Clinical staff include former drug users who themselves were resocialized in TCs. For both staff and clients, “The essential dynamic in the TC is mutual self-help” (De Leon, 1986). Through repetition and reinforcements as part of the daily regimen of community roles and community life, the TC process seeks to create a new drug-free lifestyle for the client.

Throughout the 1960s and into the 1970s, TCs were the most visible type of treatment in correctional settings. The first was established in 1962 in the Nevada State Prison, and in 1967 the Federal Bureau of Prisons established a similar program in an institution in Danbury, Connecticut (Petersen, 1974). During the next decade, many prison-based TCs seemed to come and go—in Arkansas, Connecticut, Georgia, Michigan, Nebraska, New York, Okla- homa, South Carolina, Virginia, the federal system, and elsewhere (Smith, 1977; Toch, 1980; Weppner, 1983). By the early 1980s, almost all of these TCs had closed, the result of prison crowding, state budget deficits, staff burnout or blowup, and changes in prison leadership. In a few instances, shut-downs were precipitated by inmate residents smuggling drugs and alco- hol into the TC units. In others, custodial officers distrusted TC staff and operations and deliberately sabotaged programs (Camp & Camp, 1990).

Despite these early failures, the TC reemerged in the late 1980s and dem- onstrated great promise in the treatment of substance abuse (De Leon, 1985; Yablonsky, 1989). TCs modified their approaches from the earlier Synanon models. Staff had more professional training and credentials, inmates were given appropriate power and rewards without too much program control, and TC programs worked with institutional staff rather than in opposition (Wexler, 1995). At the same time, correctional approaches were again becoming more favorable to the concept of rehabilitation, no doubt spurred by the massiveprison crowding accompanying the increased use of determinate and mandatory

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 297

minimum sentences for drug-involved offenders (Inciardi & Martin, 1993; Wexler, 1995). Moreover, correctional administrators and clinicians found TCs useful in improving client attitudes and inmate comportment (Field, 1989; Holland, 1978; Inciardi & Scarpitti, 1992; Wexler, Falkin, Lipton, Rosenblum, & Goodloe, 1988).

A CONTINUUM OF TC TREATMENT

More recently, Inciardi and colleagues have argued that an integrated con- tinuum of corrections-based TC treatment works best for seriously drug- involved offenders. This continuum involves three stages of TC treatment, tied to an inmate’s changing correctional status: prison→ work release→ parole or other form of community supervision (Inciardi, Lockwood, & Mar- tin, 1991, 1994).

The primary stage of treatment should consist of a prison-based TC designed to facilitate personal growth through the modification of deviant lifestyles and behavior patterns. Segregated from the rest of the penitentiary, recovery from drug abuse and the development of prosocial values in the prison TC would involve essentially the same mechanisms seen in community-based TCs. Ideally, it should last for 9 to 12 months (Wexler et al., 1988), and client recruits should be within 12 to 15 months of work release eligibility.

Since the 1970s, work release has become a widespread correctional prac- tice for felony offenders. It is a form of partial incarceration where inmates work for pay in the free community but must spend their nonworking hours in an institution or a community-based work release facility or halfway house (Inciardi, 1999). Inmates qualified for work release are those approaching their parole eligibility or conditional release dates. Although graduated release of this sort carries the potential for easing an inmate’s process of com- munity reintegration, there is a negative side, especially for those whose drug involvement served as their entry key to the penitentiary gate in the first place. This initial freedom exposes inmates to old groups and behaviors that can easily lead them back to substance abuse, criminal activities, and reincar- ceration. Even those receiving intensive TC treatment while in the institution face the prospect of their recovery breaking down. Work release environ- ments often do little to stem the process of relapse. Because work releasees mirror the institutional populations from which they came, they often con- tinue the negative values of the prison culture. Added to the continuing inmate culture is the new (albeit limited) freedoms of being able to go out and work in the community without continual supervision. In a correctional

298 THE PRISON JOURNAL / September 1999

setting where street drugs and street norms tend to abound, it is not surprising that many relapse or recidivate.

Graduates of prison-based TCs are at a special disadvantage in a tradi- tional work release center, because they must live and interact in what is typi- cally an antisocial, nonproductive setting. Without clinical management and proper supervision, their recovery can be severely threatened. Thus, secon- dary TC treatment is warranted. This secondary stage is a transitional TC—the therapeutic community work release center—with a program com- position similar to that of the traditional TC.

In the tertiary or aftercare stage, clients have completed work release and are living in the free community under parole or some other form of supervi- sion. Treatment intervention in this stage should involve continued monitor- ing by TC counselors including regular outpatient counseling, group therapy, and family sessions.

THE DELAWARE TC CONTINUUM

The use and practicality of such a continuum of correctional TC treatment was untested until only recently. The three-stage model has been put in place sequentially in the Delaware correctional system. Initially, it was built around two therapeutic communities—the KEY (a prison-based program for men begun in 1988) and CREST Outreach Center (a work release program for both men and women begun in 1991). Subsequently, and largely based on the initial success of the KEY and CREST, Delaware’s TC resources have been expanded by the addition of three new prison TCs and two community- based transitional facilities.

The unique contribution in Delaware is CREST Outreach Center, the work release therapeutic community program in Wilmington, Delaware (Lockwood, Inciardi, & Surratt, 1997). CREST emerged as the clinical com- ponent of a NIDA treatment research demonstration project, awarded to the University of Delaware in 1990. The primary goal of the research was to examine the feasibility and clinical efficacy of a therapeutic community work release center for drug-involved felony offenders who had spent a number of years in the penitentiary. The issues of feasibility and efficacy were especially important because CREST represented the first attempt, any- where in the world, at developing a correctional work release program built on a therapeutic community model.

CREST Outreach Center was designed to provide the secondary or transi- tional phase in the continuum of treatment and to eventually serve as a home base for developing a tertiary aftercare program. As such, clients from the

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 299

in-prison TC would transition to CREST. However, it was also planned to use CREST to examine the effects of work release TC treatment for clients who did not come from a prison TC. Consequently, not all of CREST’s residents receive their primary treatment in prison; in fact, for many, their first expo- sure to a TC is at CREST. So, the Delaware work release TC environment at CREST represents secondary treatment for some (those coming from the prison TC), but for even more clients, CREST is both their primary and sec- ondary treatment.

METHOD

RESEARCH DESIGN

The research protocol includes ongoing process evaluation and baseline and multiple follow-up interviews with all treatment and comparison clients, as well as HIV and urine testing at each contact. The baseline interview is administered in prison, prior to an inmate’s transfer to CREST Outreach Center or regular work release. The first follow-up occurs 6 months hence, corresponding with graduation from CREST (for the treatment groups) or completion of regular work release (for the control group). Subsequent inter- views have been conducted 18 months and 42 months after baseline (i.e., 1 year and 3 years after treatment and work release). Treatment dropouts are also followed. Interviews at baseline and each subsequent follow-up are lengthy, representing 700 variables per administration, including data on basic demographics, living situations, criminal history, drug use history, treatment history, sexual behavior history, sexual attitudes, HIV risks, self- esteem, sensation seeking, and physical and mental health. The instruments include much of the Addiction Severity Index (ASI) and NIDA’s Risk Behav- ior Assessment (RBA). It is important to note that these instruments were administered by the researchers after client selection and not as part of the client selection process. Follow-up surveys elicit detailed event history infor- mation on the intervening periods. Participation in the project is voluntary, and participants are paid up to $50 at each of the testing intervals—$25 for completing the questionnaire and $25 for giving both a blood and urine sam- ple. About 95% of all eligible subjects agreed to participate in the study at baseline, and more than 80% of those interviewed also provided a blood and/or urine specimen. Follow-up rates for all study participants have been about 80%.

The original research project design to evaluate the TC continuum in Delaware was quasi-experimental, and it has involved a number of both

300 THE PRISON JOURNAL / September 1999

planned and serendipitous comparisons of prison releasees. Most of the pre- viously published outcome work focused on four treatment and control groups (Inciardi, 1996; Inciardi et al., 1997; Martin et al., 1995). The first group is a sample of conventional work release residents with a past history of drug use, as determined by the Delaware Department of Corrections coun- seling and institutional review procedures. The department’s classification is based on criminal history and correctional counselor interviews; counselor recommendations are approved by the Institutional Classification Board as part of the classification from prison to work release. Among those classified to work release with a recommendation (not a mandate) for drug treatment, some were randomly assigned to regular work release (the comparison group). These releasees have relative freedom during working hours, but are held in secure dormitories after 10 p.m. Most attend Alcoholics Anonymous (AA) or Narcotics Anonymous (NA) meetings at the work release center and have access to an on-site counselor, but have little other treatment. The quasi-experimental contrast in the study is between this comparison group and the second group, the CREST clients, a sample of work releasees with a history of drug use and a recommendation for treatment who were randomly assigned to CREST. This was possible because the number of eligible clients exceeded the capacity at CREST to allow for the random assignment proce- dure. So, subjects coming to work release with a history of chronic, heavy drug use (determined by institutional classification screening) but no prison TC experience were randomly assigned to one of these two groups. The assignments were made by the research staff and not by clinical staff from the population of those classified to work release with a history of significant drug problems.

Two other groups also are being followed as part of this study: (a) the KEY group, a group of releasees from the in-prison TC who did not go to CREST because they were released before CREST was operational; and (b) the KEY-CREST group, clients who graduated from the KEY and were then assigned to CREST for work release treatment. Although not randomly assigned, each of these groups included all clients coming from the KEY who were being classified to work release. More importantly, each of these nonrandom groups provides an important contrast with the randomly assigned groups. Table 1 presents some of the salient distinctions among the four initial comparison groups, which we refer to as initial groups.

These initial group distinctions are conservative in that they consider assignment to CREST rather than whether the client completed treatment. A number of clients who enter CREST do not stay more than 1 month, and oth- ers stay an appreciable time but do not complete the program. In practice,

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 301

CREST completion rates are about 65%. Consequently, our initial group categories do not account for the potential cumulative effect of length of time in transitional TC treatment, nor do they consider the potential categorical effect of completing and graduating from CREST. There are other limitations of the initial groups in operationalizing the TC continuum. The initial groups could not be used to evaluate the effectiveness of the full theoretical contin- uum of treatment, because the tertiary or aftercare program did not exist in Delaware until 1996. Aftercare has now been in operation long enough to have a first cohort of aftercare clients who are 3 years from their graduation from work release. At the other end of the continuum, the in-prison TC-only group (KEY) has become a historical artifact because almost all KEY gradu- ates are assigned to CREST for their work release assignment. Finally, as noted earlier, for the majority of TC clients for whom we have long-term follow-up data, CREST represents both primary and secondary treatment.

Consequently, we have begun to examine a different four-group compari- son, which we refer to in this article as revised groups. These include com- parison (the same group as previously described), CREST dropouts, CREST graduates without aftercare, and CREST graduates with aftercare. This grouping more appropriately distinguishes major elements of the TC contin- uum, particularly in the secondary and tertiary stages that have not been ade- quately studied. Also, and significantly, we have excluded the KEY-only group from the revised groups, and we have included KEY graduates who went to CREST together with those who went directly to CREST. Table 2 describes the distinctions for these four revised groups.

302 THE PRISON JOURNAL / September 1999

TABLE 1: Characteristics of Initial Groups in the Delaware Therapeutic Community (TC) Continuum for Subjects About to be Released From Prison

Assigned Work Past In-Prison to CREST

Release Heavy Randomly TC (KEY) Work Classified Drug Use Gender Assigned Graduate Release TC

Comparison Yes Yes 80% male, Yes No No 20% female

KEY Yes Yes 100% male No Yes No CREST Yes Yes 80% male, Yes No Yes

20% female KEY-CREST Yes Yes 100% male No Yes Yes

initially, but decreasing to

75% male, 25% female

SAMPLE AND MEASURES

All baseline measures are self-report items. The baseline variables were chosen to represent basic demographic characteristics, and to control for potential differences in clients in drug and criminal justice histories. Cate- gorical baseline measures include gender, previous drug treatment, and eth- nic group. Previous drug use is an ordinal scale derived from questions ask- ing frequency of use of each of injecting or noninjecting cocaine, heroin, speed, crack, PCP, hallucinogens, and nonprescribed sedatives, stimulants, tranquilizers, analgesics, or other opiates in the 6 months prior to incarcera- tion. The maximum reported use of any drug was recorded on a scale ranging from 0 (no use) to 6 (use more than once a day). Continuous baseline mea- sures are age, age of first incarceration, number of previous incarcerations, and number of previous arrests. Table 3 presents these baseline variables by the initial and revised groupings.

As seen in the top half of Table 3 and as would be expected from the assignment process in the initial groups, the CREST and comparison groups are very similar but differ from the KEY groups on several dimensions. The two KEY groups contain more African Americans and were first incarcer- ated at a younger age. They were more drug-involved before entering prison. The KEY-only group is historically all male. Everyone from the KEY groups had previous drug abuse treatment, whereas more than one fifth of those in the other two groups reported no prior treatment.

The 6-month follow-up interviews provided a measure of time spent in any drug treatment since release from prison. This is important because the

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 303

TABLE 2: Characteristics of Revised Groups in the Delaware Therapeutic Community (TC) Continuum for Subjects About to be Released From Prison

Assigned Work Past In-Prison to CREST

Release Heavy Randomly TC (KEY) Work Classified Drug Use Gender Assigned Graduate Release TC

Comparison Yes Yes 80% male, No No No 20% female

CREST Yes Yes 80% male, No Some Yes dropouts 20% female

CREST Yes Yes 80% male, No Some Yes completers 20% female

CREST Yes Yes 80% male, No Some Yes completers 20% female with aftercare

TABLE 3: Baseline Sample Characteristics by Initial Grouping and Revised Grouping: Delaware Therapeutic Community Continuum

Initial Groups Revised Groups

KEY- CREST CREST CREST Completers Comparison KEY CREST CREST Comparison Dropouts Completers With Aftercare

n 165 38 157 68 210 109 101 69 Age (M) 29 31 29 31 29 29 31 31 Age of first incarceration 22 18 21 19 22 20 21 23 Mean number of times in prison 4 4 4 4 4 4 4 4 Mean number of arrests 10 15 12 11 10 13 13 11 Males (%) 82 100 77 74 82 83 76 67 Whites (%) 27 16 24 13 28 21 15 20 Hispanic (%) 4 0 2 0 3 4 2 2 African Americans (%) 68 84 74 85 68 74 83 74 Drug use 6 months prior to prison,

ranging from 0 (none) to 6 (several times a day) 4 5 4 5 4 4 5 4

Previous drug treatment (%) 75 100 77 100 74 79 79 88

3 0

4

comparison sample was not truly a no treatment group. As might be expected in the real world, many of these releasees were motivated for treatment and sought treatment on their own during work release. Beyond the ethical issue of denying treatment to those seeking it, the subsequent follow-up data reveal the practical observation that fully 56% of the no treatment group reported some treatment during the initial 180 days after prison release (the time while the CREST group was at the TC work release program). The mean number of days in treatment was, not unexpectedly, higher for the CREST assignees (116 days), as contrasted with 24 days for the comparison group. Because of the skew in number of days, the log days of treatment was used in the follow- ing analyses.

The dependent variables for the analyses are dichotomous measures of behaviors—relapse and recidivism. Each dependent variable combines information from repeated self-report and objective criteria. The dependent measures are stated positively. The criterion for “drug-free” is whether the respondent reported no use of any of the drugs counted in the baseline mea- sure and tested negative for any illegal drug at any time in the follow-up period. In this article, we examine findings from interviews 1 year and 3 years after leaving work release. If the respondent reported no illegal drug use and tested negative, drug status was coded as 1 for drug-free; any positive response or test result was coded as 0. As such, this is an extremely conserva- tive criterion. It combines repeated self-reports and repeated urine screens during follow-up, and relapse on even just one occasion during follow-up is coded as 0. The criterion for “no arrest” included no self-reports of arrest or records of arrest for new offenses since release from prison. It does not include parole violations. Arrest status is coded 1 for no arrest for any new crime, and 0 if there is any arrest for a new crime.

RESULTS

In our earlier work, the basic hypothesis was that drug-involved offenders receiving primary treatment in a prison-based TC, followed by secondary treatment in a work release TC, would be more likely to remain arrest-free and drug-free than those having had fewer stages of TC treatment. This work examined outcomes 1 year after work release (Inciardi et al., 1997). The data were first analyzed as a bivariate logistic regression with group assignment as the only independent variable, modeled by a dummy classification with the comparison group as the excluded category. These results were strongly supportive of the continuum of TC treatment hypothesis, with each of the treatment groups showing significant and incremental improvements in

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 305

arrest-free and drug-free status during the 1-year follow-up. Moreover, the pattern of results remained similar when a number of other putative predic- tors of treatment effect—the covariates described above—were included in the analyses in multivariate logistic regressions. These control variables are not only potential predictors of treatment outcome on their own, but, more importantly, they are factors that may differ in the groups who were not part of the random assignment process (see Table 1). Also included in the model was length of time in treatment (log of number of days from 0-180) during the 6 months of work release to demonstrate the TC program effect beyond retaining clients in treatment longer.

These previous results reported in 1997 and predicting arrest-free and drug-free status are reviewed in Figure 1. The results of the multivariate logistic regressions predicting each of the two dichotomous arrest-free and drug-free criterion, including the effects of these other independent vari- ables, are presented as logistic regression histograms, showing the predicted percentages (or probabilities) for each comparison group. This was done by estimating 1/(1 +e–Z), whereZ is the sum of the logistic regression equation for each group with other variable effects entered as their coefficient multi- plied by the mean value for that variable.

In examining the prediction of each arrest-free and drug-free status, the comparison group did not differ significantly from the KEY group. As reported earlier (Inciardi et al., 1997), the data suggest that in-prison TC pri- mary treatment alone may not be effective. More effective is the provision of treatment in the transitional setting (CREST), where treatment occurs at the time that clients are truly at risk for relapse and new criminal activity. The KEY-CREST group, with both primary and secondary TC treatment, emerges as the group doing by far the best in terms of avoiding relapse and recidivism.

This earlier work demonstrated a consistent and persuasive pattern of results supporting the continuum of TC treatment in a 1-year follow-up study. It is, of course, not strictly appropriate to apply standard tests of sig- nificance to both random and purposive comparison groups in the same analyses, and it would be particularly inappropriate without controls for known group differences. However, making such comparisons is often nec- essary in field evaluation studies. When conducted with sufficient samples, sufficient control variables, and repeated follow-ups, such studies can dem- onstrate convincing effects and have important policy implications.

Even more important from a policy perspective is whether the effects of TC treatment on drug-involved offenders have long-term effects. Significant reductions in relapse and recidivism 1 year after work release have substan- tial societal benefit, but, if the effects are longer lived, the cost savings to

306 THE PRISON JOURNAL / September 1999

society increase almost exponentially. Consequently, we proceed to look at these same criteria of arrest-free and drug-free—for these same groups and with the same control variables used in the 1997 analyses—for a period of 3 years after work release. The research sample attrition, fortunately, has been low (less than 10% between the 1-year and 3-year follow-up), and, signifi- cantly, it has not varied appreciably between TC clients and the comparison group. Again we estimate a multiple logistic regression model for each

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 307

Figure 1: Percentage Arrest-Free and Drug-Free for Initial Groups in Year After Release (adjusted for control variables)

*Significantly different from the comparison group, p < .05.

dichotomous dependent variable. The results of this regression predicting arrest-free at 3 years are reported in Table 4, and the results for drug-free at 3 years are reported in Table 5. Each table includes the regression coefficients, significance levels, and Wald statistics for each of the independent variables. We first review the effects of the control variables and then focus on the dif- ferences among the initial groups.

In general, the effects of the control variables in the multivariate logistic equations were modest but did indicate the importance of controlling for covariates where possible. In Table 4, the model predicting arrest-free status, the only significant covariates were age and number of previous arrests. Not surprisingly, older subjects were less likely to be arrested for a new crime, whereas those with more past arrests were more likely to have a new arrest. What is perhaps surprising is that none of the other measures—gender, race (represented here as a dichotomy of African American vs. other), early incar- ceration, previous drug use, or previous treatment—predicted new arrests. In Table 5, the model predicting drug-free status, the only significant control variable was previous drug use. These results are very similar to what had been found previously for the control variables at 1 year: very modest effects with past arrests predictive of future arrests and past drug use predictive of future drug use.

308 THE PRISON JOURNAL / September 1999

TABLE 4: Logistic Regression Coefficients for Variables Included in the Initial Group Model Predicting Arrest-Free at 3 Years

Independent Regression Standard Wald Variable Coefficient Error Statistic Significance

Age –0.062 0.022 7.629 .006 Age of first incarceration –0.042 0.025 2.834 .092 Number of times in prison 0.054 0.075 0.527 .468 Number of arrests 0.044 0.018 6.201 .013 Female = 1, male = 0 –0.279 0.300 0.862 .353 African American = 1, other = 0 –0.017 0.292 0.003 .954 Drug use 6 months prior to

prison, ranging from 0 (none) to 6 (several times a day) 0.074 0.070 1.131 .288

Previous drug treatment at baseline (1 = yes, 0 = no) –0.261 0.313 0.696 .404

Log days in treatment during work release –0.035 0.090 0.155 .694

KEY (group = 1, other = 0) –0.491 0.508 0.932 .334 CREST (group = 1, other = 0) –0.315 0.388 0.659 .417 KEY-CREST (group = 1,

other = 0) –0.571 0.447 1.631 .202 Constant 2.920 0.824 12.558 .000

NOTE: –2 log likelihood = 420.54.

What is different in the 3-year data results is the initial group effects. The coefficients for the initial group differences as deviations from the omitted comparison group and adjusted for the control variables are reported near the bottom of Tables 4 and 5. They are also shown in Figure 2 as adjusted logistic histograms. The model is exactly the same as that in Figure 1, with the same initial group comparisons, except that in Figure 2 the period at risk is now 3 years rather than 1 year.

It is readily apparent that the significant group differences in effects of TC treatment seen in Figure 1 for the initial groups have been significantly attenuated. For arrest-free status, there is a small, now nonsignificant, benefit of TC treatment versus the comparison group, and no differences among the three TC groups. For drug-free status, TC treatment in each of the three cate- gories does produce a significant increase in the probability of being drug- free, as opposed to the comparison group. However, among the TC groups, there are no differences in likelihood of being drug-free.

The implications of these findings are, at first glance, sobering. There is a modest benefit for some TC treatment as compared to no TC treatment, but no indication of the relative improvement with more stages of TC treatment, as appears so convincingly in the 1-year follow-up data. However, as we noted earlier, the initial group distinctions for each of CREST and the

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 309

TABLE 5: Logistic Regression Coefficients for Variables Included in the Initial Group Model Predicting Drug-Free at 3 Years

Independent Regression Standard Wald Variable Coefficient Error Statistic Significance

Age –0.025 0.028 0.760 .383 Age of first incarceration –0.001 0.031 0.000 .988 Number of times in prison 0.085 0.096 0.775 .379 Number of arrests –0.019 0.018 1.019 .313 Female = 1, male = 0 0.102 0.407 0.063 .802 African American = 1, other = 0 –0.256 0.374 0.470 .493 Drug use 6 months prior to

prison, ranging from 0 (none) to 6 (several times a day) 0.221 0.087 6.406 .011

Previous drug treatment at baseline (1 = yes, 0 = no) 0.129 0.411 0.098 .754

Log days in treatment during work release –0.033 0.122 0.074 .785

KEY (group = 1, other = 0) –1.843 0.677 7.419 .006 CREST (group = 1, other = 0) –1.628 0.569 8.173 .004 KEY-CREST (group = 1,

other = 0) –1.656 0.641 6.678 .010 Constant 2.664 1.028 6.716 .010

NOTE: –2 log likelihood = 291.23.

KEY-CREST were conservative in that they included all clients who entered CREST, not just CREST completers. So the models were not truly testing the continuum hypothesis.

310 THE PRISON JOURNAL / September 1999

Figure 2: Percentage Arrest-Free and Drug-Free for Initial Groups in 3 Years After Release (adjusted for control variables)

*Significantly different from the comparison group, p < .05.

Therefore, given the limitations of the initial groups and the recent exis- tence of sufficient cases with an aftercare program following CREST, we revised our comparative grouping to look at three more realistic distinctions of engagement in the TC continuum—CREST dropouts, CREST completers (without aftercare), and CREST completers with aftercare—as opposed to a comparison group with no TC treatment. In essence, the revised groups allow us to examine the effects of secondary and tertiary TC treatment. We ignore the primary stage distinction for the moment out of necessity. Of the 279 CREST clients included in the analyses, only 21% have come from an in- prison TC. However, coming from an in-prison TC does seem predictive of retention in the TC continuum: 14% of the CREST dropouts, 21% of the CREST completers, and 32% of the CREST completers with aftercare came from an in-prison TC.

In the final models presented, we again look at the criteria of arrest-free status and drug-free status with the same control variables for a period of 3 years after work release, but for the revised grouping of comparison, CREST dropouts, CREST completers, and CREST completers with aftercare. The regression coefficients, significance levels, and Wald statistics are reported in Table 6 for arrest-free status and Table 7 for drug-free status. The revised group differences, adjusted for the control variables, are then shown in Fig- ure 3 as adjusted logistic histograms.

Looking at the control variable effects first, the results are very similar to what was observed in the initial groups models. As seen in Table 6, age and previous arrest history again predict new arrests, but here there is also a small effect of previous drug history. This may indicate that drug history is related to having gone to a prison TC, a possibility indicated by the higher scores on prior drug history for the KEY groups (see Table 3). In Table 7, the model predicting drug-free status for the revised groups, the only significant control variable is prior drug history.

Whereas the control variable effects are quite stable in each of the initial group and revised group models, the group effects themselves are quite dif- ferent. The top panel of Figure 3 (predicting arrest-free) reveals that CREST dropouts are as likely to be arrested on a new charge as the comparison group. However, those who complete CREST do much better, and those who com- plete CREST and receive aftercare as well are the least likely to have a new arrest. Less than one out of three clients with aftercare have a new arrest, whereas more than two out of three of the comparison group with no TC treatment have a new arrest. It is new arrests and the concomitant sentence likely associated with a new crime by a repeat offender (because all clients in the study are repeat offenders) that have the most obvious implications for the cost savings of treatment.

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 311

A stepping stone pattern for the secondary and tertiary treatment groups is even more apparent when drug-free status is examined in the lower panel of Figure 3. When contrasted with the comparison group, CREST dropouts are more than three times as likely to be drug-free, CREST completers more than five times as likely, and CREST completers with aftercare are seven times more likely to be drug-free. It appears that length of time in transitional and aftercare TC treatment is a significant and incremental protective factor against likelihood of relapse.

DISCUSSION

In our previous work, we have suggested the use of the TC approach to treatment for drug-involved offenders. Given the usually long-standing drug and criminal histories of these clients, treatment needs to be both intensive and extensive, and we have proposed a continuum of primary (in-prison), secondary (work release), and tertiary (aftercare) TC treatment correspond- ing to sentence mandates. We have been fortunate in Delaware, with federal and state support, to institute such a continuum in stages over the past decade.

312 THE PRISON JOURNAL / September 1999

TABLE 6: Logistic Regression Coefficients for Variables Included in the Revised Group Model Predicting Arrest-Free at 3 Years

Independent Regression Standard Wald Variable Coefficient Error Statistic Significance

Age –0.055 0.021 6.984 .008 Age of first incarceration –0.034 0.022 2.256 .133 Number of times in prison –0.055 0.061 0.803 .370 Number of arrests 0.027 0.012 5.487 .019 Female = 1, male = 0 –0.350 0.281 1.552 .213 African American = 1, other = 0 –0.030 0.269 0.012 .912 Drug use 6 months prior to

prison, ranging from 0 (none) to 6 (several times a day) 0.121 0.059 4.215 .040

Previous drug treatment at baseline (1 = yes, 0 = no) –0.179 0.279 0.412 .521

Log days in treatment during work release 0.071 0.082 0.766 .381

CREST dropouts (group = 1, other = 0) 0.034 0.385 0.008 .930

CREST completers (group = 1, other = 0) –1.127 0.390 8.363 .004

CREST completers with aftercare (group = 1, other = 0) –1.711 0.426 16.152 .000

Constant 2.703 0.737 13.453 .000

NOTE: –2 log likelihood = 462.90.

Earlier evaluations that we conducted of the Delaware programs demon- strated significant reductions in relapse and recidivism for offenders who received primary and secondary or just secondary TC treatment, as opposed to a no-treatment comparison group or a group that received only in-prison treatment. These initial-group differences remain clearly evident 12 months after leaving treatment and work release (18 months after release from prison), an outcome period longer than most reported in the literature (see Figure 1). The importance of providing primary and/or secondary treatment in a transitional setting was highlighted.

However, much of the effect of exposure to primary and secondary TC treatment appeared to decline significantly when the time at risk involved a period 3 years after completing treatment (see Figure 2). Declines in relapse were seriously attenuated and, in fact, declines in recidivism for those enter- ing treatment almost entirely disappeared. Earlier reported effects of the con- tinuum of treatment were also no longer present. These initial groups, how- ever, were overly conservative, and do not effectively operationalize the continuum model because the CREST and KEY-CREST initial groups included all those assigned to CREST without consideration of whether they

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 313

TABLE 7: Logistic Regression Coefficients for Variables Included in the Revised Group Model Predicting Drug-Free at 3 Years

Independent Regression Standard Wald Variable Coefficient Error Statistic Significance

Age –0.000 0.027 0.000 .986 Age of first incarceration –0.019 0.029 0.416 .519 Number of times in prison 0.019 0.080 0.056 .813 Number of arrests –0.014 0.012 1.282 .258 Female = 1, male = 0 0.155 0.360 0.186 .666 African American = 1, other = 0 –0.196 0.342 0.328 .567 Drug use 6 months prior to

prison, ranging from 0 (none) to 6 (several times a day) 0.219 0.069 9.982 .002

Previous drug treatment at baseline (1 = yes, 0 = no) –0.073 0.369 0.039 .844

Log days in treatment during work release 0.007 0.115 0.004 .950

CREST dropouts (group = 1, other = 0) –1.280 0.534 5.743 .017

CREST completers (group = 1, other = 0) –1.841 0.552 11.112 .001

CREST completers with aftercare (group = 1, other = 0) –2.245 0.554 16.401 .000

Constant 2.574 0.922 7.784 .005

NOTE: –2 log likelihood = 317.19.

completed or dropped out of the transitional TC. Moreover, the essential component of aftercare did not exist in the program. Under these circum- stances, it is more surprising that the 1-year effects are so strong, rather than that the 3-year effects weaken. The 3-year outcome becomes far more

314 THE PRISON JOURNAL / September 1999

Figure 3: Percentage Arrest-Free and Drug-Free for Revised Groups in 3 Years After Release (adjusted for control variables)

*Significantly different from the comparison group, p < .05.

positive when we examine the revised groups where program participation, program completion, and aftercare are taken into account. The multivariate logistic regression results for the revised groups (see Figure 3) suggest that clients who complete secondary treatment (some of whom also completed primary treatment) do better than those with no treatment or program drop- outs after 3 years. Furthermore, data now available on Delaware clients who received tertiary treatment (a TC aftercare program) suggest that this group does even better still in predicting the long-term likelihood of remaining drug- and arrest-free. The results provide new and long-term support for the continuum model of TC treatment for seriously drug-involved offenders.

These results are still preliminary. As we gain 3-year follow-up data on more cases, we will be able to look comprehensively at all stages and most combinations of the three-stage continuum of TC treatment. In particular, length of time in each stage of treatment as well as completion of each stage can be examined more fully. So far we have looked at primary/secondary stage effects and now secondary/tertiary stage effects. We have not been able to put the whole model together. The new analyses do not distinguish the in- prison TC experience effects, although we noted that there is some indication that in-prison TC graduates are more likely to remain in treatment through work release and aftercare. With more clients having 3-year follow-up, it will be possible to include all stages of the continuum in one model.

There are other limitations that need to be addressed in future research. It is obvious that the models estimated, though significant, are not accounting for all of the variance in predicting relapse and recidivism. It is probable that important control variables and possible that significant confounding vari- ables for group effects have not been modeled. One large potential area that needs to be considered is the selection of clients into treatment and the suit- ability of the treatment that clients receive. This article has ignored the exten- sive literature on assessing client risk and needs and getting them into appro- priate programs (e.g., Andrews et al., 1990; De Leon et al., 1995; De Leon, Melnick, Kressel, & Jainchill, 1994). There is a less extensive but equally important literature on compulsory versus voluntary treatment (De Leon et al., 1995; Leukefeld & Tims, 1988). Both of these issues are involved in the Delaware data and need to be operationalized and incorporated in any com- prehensive model involving treatment effects predicting relapse and recidivism.

Despite these limitations, the present data support the value of treatment in work release and parole settings and the importance of retention in treat- ment in predicting long-term success in reducing the likelihood of recidivism or relapse. More generally, the data also support some long-standing argu- ments about the value of aftercare (De Leon, 1990-1991; Inciardi & Scarpitti,

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 315

1992). In 1935, for example, just 1 year after the termaddiction first appeared in the American Psychiatric Association’s Standard Classified Nomenclature of Disease, the U.S. Public Health Service opened its first nar- cotics “farm” in Lexington, Kentucky. A second facility was opened in Fort Worth, Texas, 3 years later. They were called farms because clients partici- pated in farm work during their sojourns at Lexington and Fort Worth. In reality, however, both facilities were prison hospitals established for the treatment of addiction to narcotics while at the same time designed to allevi- ate prison crowding in other parts of the federal system (White, 1998, p. 122).

It was anticipated from the outset that the treatment approach at the new federal establishments would be highly effective, for at the very least, the hospitals were designed to treat not only the physical dependence but also the mental and emotional problems thought to be related to addiction (Maddux, 1978). This was an advanced conception, because until then treatment for a narcotics problem had focused exclusively on physical dependence. And there were other innovations, including a drug-free environment and access to educational and vocational services, as well as recreational and religious activities. When the patients were followed up, however, treatment outcomes were disappointing. In a study of 1,881 patients discharged from the Lexing- ton facility to the New York City area, for example, some 90% had become readdicted within 1 to 4½ years (Hunt & Odoroff, 1962). Other studies docu- mented that 90% to 96% treated at Lexington returned to active addiction, most within 6 months of discharge (White, 1998, p. 125). Although Dr. Victor H. Vogel, the medical officer in charge of the Lexington program, tried to put a positive spin on the outcome data by suggesting that the treatment results were better than those of such other chronic diseases as diabetes and cancer (Maddux, 1978), the studies underscored the limited role of institutionally based treatment alone in reducing the likelihood of subsequent relapse. What the Lexington and Fort Worth experiences did suggest was the need for community-based treatment and a continuum of treatment with community aftercare as a crucial component (Brown, 1979).

The data presented in this article present some interesting parallels and comparisons with the Lexington and Fort Worth experiences. When the KEY was established in the Delaware correctional system in 1988, the authors argued that unless a continuum of treatment from prison to work release to aftercare was established, the positive effects of the institutional phase of treatment could not be maximized. This prediction is amply documented in Figure 1, wherein there were no significant differences in relapse and recidi- vism rates between the KEY and comparison groups at the 1-year follow-up. Prison treatment alone does not appear to have a lasting impact. Significant differences after 1 year are readily apparent, however, as care extends into the

316 THE PRISON JOURNAL / September 1999

community during the work release transition. Next, a comparison of Figures 2 and 3 shows how important it is to take into account completing transitional TC treatment at CREST. Finally, and as illustrated in Figure 3, the most potent effects of treatment become visible when followed by community aftercare. These data should not suggest, however, that prison-based treat- ment is without merit. On the contrary, when combined with transitional treatment in a work release setting followed by aftercare, the positive effects become stronger. As follow-ups continue and the number of cases becomes larger, the effects of prior treatment—both in the institution and else- where—can be more fully examined.

REFERENCES

Andrews, D. A., Zinger, I., Hoge, R. D., Bonta, J., Gendreau, P., & Cullen, F. T. (1990). Does correctional treatment work? A clinically relevant and informed meta-analysis.Criminol- ogy, 28(3), 369-404.

Apsler, R. (1991). Evaluating the cost-effectiveness of drug abuse treatment services. In W. S. Cartwright & J. M. Kaple (Eds.),Economic costs, cost-effectiveness, financing, and community-based drug treatment(NIDA research monograph 113, pp. 57-66). Rockville, MD: U.S. Department of Health and Human Services.

Apsler, R., & Harding, W. M. (1991). Cost-effectiveness analysis of drug abuse treatment: Cur- rent status and recommendations for future research. In National Institute on Drug Abuse, Drug abuse services research series, No. 1: Background papers on drug abuse financing and services approach. Washington, DC: Government Printing Office.

Auerbach, J. D., Wypijewska, C., & Brodie, K. H. (1994).AIDS and behavior. Washington, DC: National Academy Press.

Brown, B. S. (1979).Addicts and aftercare: Community integration of the former user. Beverly Hills, CA: Sage.

California Department of Alcohol and Drug Programs. (1994).Evaluating recovery services: The California Drug and Alcohol Treatment Assessment (CALDATA). Sacramento, CA: Author.

Camp, G. M., & Camp, C. G. (1990).Preventing and solving problems involved in operating therapeutic communities in a prison setting. South Salem, NY: Criminal Justice Institute.

Chaiken, M. R. (1989).In prison programs for drug involved offenders. Washington, DC: Office of Justice Programs, National Institute of Justice.

Currie, E. (1998).Crime and punishment in America. New York: Henry Holt. De Leon, G. (1985). The therapeutic community: Status and evolution.International Journal of

the Addictions, 20, 823-844. De Leon, G. (1986). The therapeutic community for substance abuse: Perspective and approach.

In G. De Leon & J. T. Ziegenfuss (Eds.),Therapeutic communities for addictions: Readings in theory, research and practice(pp. 5-18). Springfield, IL: Charles C Thomas.

De Leon, G. (1990-1991). Aftercare in therapeutic communities.International Journal of the Addictions, 25, 1225-1237.

De Leon, G., Inciardi, J. A., & Martin, S. S. (1995). Residential drug treatment research: Are conventional control designs appropriate for assessing treatment effectiveness?Journal of Psychoactive Drugs, 27(1), 85-91.

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 317

De Leon, G., Melnick, G., Kressel, D., & Jainchill, N. (1994). Circumstances, motivation, readi- ness and suitability (the CMRS scales): Predicting retention in therapeutic community treat- ment.American Journal of Drug and Alcohol Abuse, 20, 495-515.

Field, G. (1989, December). The effects of intensive treatment on reducing the criminal recidi- vism of addicted offenders.Federal Probation, 51-56.

Fourcier, M. W. (1991). Substance abuse, crime and prison-based treatment: Problems and pros- pects.Sociological Practice Review, 2(2), 123-131.

Graham, W. F., & Wexler, H. K. (1997). The Amity Therapeutic Community Program at Dono- van Prison: Program description and approach. In G. De Leon (Ed.),Community as method (pp. 69-86). Westport, CT: Praeger.

Harwood, H. J., Hubbard, R. L., Collins, J. J., & Rachal, J. V. (1988). The costs of crime and the benefits of drug abuse treatment: A cost-benefit analysis using TOPS data. In C. G. Leuke- feld & F. M. Tims (Eds.),Compulsory treatment of drug abuse: Research and clinical prac- tice (NIDA research monograph 86, pp. 209-235). Rockville, MD: U.S. Department of Health and Human Services.

Harwood, H. J., Napolitano, D. M., Kristianses, P., & Collins, J. J. (1984).Economic costs to so- ciety of alcohol and drug abuse and mental illness. Research Triangle Park, NC: Research Triangle Institute.

Hiller, M. L., Knight, K., & Simpson, D. D. (in press). Prison-based substance abuse treatment, residential aftercare, and recidivism.Addiction.

Holland, S. (1978). Gateway houses: Effectiveness of treatment on criminal behavior.Interna- tional Journal of the Addictions, 13, 369-381.

Hubbard, R. L., Marsden, M. E., Rachal, J. C., Harwood, H. J., Cavanaugh, E. R., & Ginzburg, H. M. (1989).Drug abuse treatment: A national study of effectiveness. Chapel Hill: Univer- sity of North Carolina Press.

Hunt, G. H., & Odoroff, M. E. (1962). Follow-up study of narcotic drug addicts after hospitaliza- tion. Public Health Reports, 77, 41.

Inciardi, J. A. (1996, June). A corrections-based continuum of effective drug abuse treatment. National Institute of Justice.

Inciardi, J. A. (1996a, June). A corrections-based continuum of effective drug abuse treatment. In National Institute of Justice research preview. Washington, DC: Department of Justice, Office of Justice Programs.

Inciardi, J. A. (1996b, October 18).Outcome evaluation of the therapeutic community contin- uum in corrections. Paper presented to the U.S. Attorney General’s Staff, Washington, DC.

Inciardi, J. A. (1999).Criminal justice(6th ed.). Ft. Worth, TX: Harcourt Brace. Inciardi, J. A., & Lockwood, D. (1994). When worlds collide: Establishing CREST Outreach

Center. In B. W. Fletcher, J. A. Inciardi, & A. M. Horton (Eds.),Drug abuse treatment: The implementation of innovative approaches(pp. 63-78). Westport, CT: Greenwood Press.

Inciardi, J. A., Lockwood, D., & Martin, S. S. (1991, May 16-17).Therapeutic communities in corrections and work release: Some clinical and policy considerations. Paper presented at the National Institute on Drug Abuse Technical Review Meeting on Therapeutic Commu- nity Treatment Research, Bethesda, MD.

Inciardi, J. A., Lockwood, D., & Martin, S. S. (1994). Therapeutic communities in corrections and work release: Some clinical and policy implications. In F. M. Tims, G. De Leon, & N. Jainchill (Eds.),Therapeutic community: Advances in research and application(NIDA research monograph 144, pp. 259-267). Rockville, MD: U.S. Department of Health and Hu- man Services.

Inciardi, J. A., & Martin, S. S. (1993). Drug abuse treatment in criminal justice settings.Journal of Drug Issues, 23(1), 1-6.

318 THE PRISON JOURNAL / September 1999

Inciardi, J. A., Martin, S. S., Butzin, C. A., Hooper, R. M., & Harrison, L. D. (1997). An effective model of prison-based treatment for drug-involved offenders.Journal of Drug Issues, 27(2), 261-278.

Inciardi, J. A., McBride, D. C., Platt, J. J., & Baxter, S. (1993). Injection drug users, incarcera- tion, and HIV: Some legal and social service delivery issues. In B. S. Brown & G. M. Besch- ner (Eds.),Handbook on risk of AIDS: Injection drug users and sexual partners(pp. 337- 351). Greenwood, CT: Greenwood Press.

Inciardi, J. A., & Scarpitti, F. R. (1992, March).Therapeutic communities in corrections: An overview. Paper presented at the annual meeting of the Academy of Criminal Justice Sci- ences, Pittsburgh, PA.

Knight, K., Hiller, M. L., & Simpson, D. D. (1999). Evaluating corrections-based treatment for the drug-abusing criminal offender.Journal of Psychoactive Drugs, 31(3), 299-304.

Knight, K., Simpson, D. D., & Hiller, M. L. (1996, August).Evaluation of prison-based treat- ment and aftercare. Paper presented at the annual meeting of the American Psychological Association, Toronto, Canada.

Leukefeld, C. G., Logan, T. K., Martin, S. S., Purvis, R. T., & Farabee, D. (1998). A health ser- vices use framework for drug-abusing offenders.American Behavioral Scientist, 41(8), 1123-1135.

Leukefeld, C. G., & Tims, F. M. (1988). Compulsory treatment: A review of the findings. In C. G. Leukefeld & F. M. Tims (Eds.),Compulsory treatment of drug abuse: Research and clinical practice(NIDA research monograph 86, pp. 236-254). Rockville, MD: U.S. Department of Health and Human Services.

Lipton, D. S. (1995).The effectiveness of treatment for drug abusers under criminal justice su- pervision. Washington, DC: National Institute of Justice.

Lockwood, D., Inciardi, J. A., & Surratt H. L. (1997). CREST outreach center: A model for blending treatment and corrections. In F. M. Tims, J. A. Inciardi, B. W. Fletcher, & A. M. Horton (Eds.),The effectiveness of innovative approaches in the treatment of drug abuse (pp. 70-82). Westport, CT: Greenwood Press.

Maddux, J. F. (1978). History of the hospital treatment program: 1935-1974. In W. Martin & H. Isbell (Eds.),Drug addiction and the U.S. Public Health Service(pp. 217-250). Washing- ton, DC: Department of Health, Education and Welfare.

Martin, S. S., Butzin, C. A., & Inciardi, J. A. (1995). Assessment of a multistage therapeutic community for drug-involved offenders.Journal of Psychoactive Drugs, 27(1), 109-116.

Mathias, R. (1995, May/June). Community treatment helps offenders stay drug and alcohol free. In NIDA notes. Rockville, MD: National Institute on Drug Abuse.

Mello, C. O., Pechansky, F., Inciardi, J. A., & Surratt, H. L. (1997). Participant observation of a therapeutic community model for offenders in drug treatment.Journal of Drug Issues, 27, 299-314.

Nielsen, A. L., & Scarpitti, F. R. (1997). Changing the behavior of substance abusers: Factors influencing the effectiveness of therapeutic communities.Journal of Drug Issues, 27(2), 279-298.

Office of Justice Programs. (1996, August).Residential substance abuse treatment program fact sheet. Washington, DC: Department of Justice.

Petersen, D. M. (1974). Some reflections on compulsory treatment of addiction. In J. A. Inciardi & C. D. Chambers (Eds.),Drugs and the criminal justice system(pp. 143-169). Beverly Hills, CA: Sage.

Rice, D. P., Kelman, S., Miller, L. S., & Dunmeyer, S. (1990).The economic costs of alcohol and drug abuse and mental illness: 1985(Report submitted to the Office of Financing and Cov-

Martin et al. / THERAPEUTIC COMMUNITY TREATMENT 319

erage Policy of the Alcohol, Drug Abuse, and Mental Health Administration, U.S. Depart- ment of Health and Human Services). San Francisco: Institute for Health and Aging.

Rouse, J. J. (1991). Evaluation research on prison based drug treatment programs and some pol- icy implications.International Journal of the Addictions, 26, 29-44.

Smith, R. (1977).Drug programs in correctional institutions. Washington, DC: Law Enforce- ment Assistance Administration.

Toch, H. (Ed.). (1980).Therapeutic communities in corrections. New York: Praeger. Weppner, R. S. (1983).The untherapeutic community: Organizational behavior in a failed ad-

diction treatment program. Lincoln: University of Nebraska Press. Wexler, H. K. (1995). The success of therapeutic communities for substance abusers in Ameri-

can prisons.Journal of Psychoactive Drugs, 27, 57-66. Wexler, H. K., De Leon, G., Thomas, G., Kressel, D., & Peters, J. (1999). The Amity prison TC

evaluation: Reincarceration outcomes.Criminal Justice and Behavior, 26(2), 147-167. Wexler, H. K., Falkin, G. P., Lipton, D. S., Rosenblum, A. B., & Goodloe, L. P. (1988).A model

prison rehabilitation program: An evaluation of the “Stay’n Out” therapeutic community. New York: Narcotic and Drug Research, Inc.

White, W. L. (1998).Slaying the dragon: The history of addiction treatment and recovery in America. Bloomington, IL: Chestnut Health Systems.

Yablonsky, L. (1989).The therapeutic community: A successful approach for treating substance abusers. New York: Gardner.

320 THE PRISON JOURNAL / September 1999