Criminal Behaviour and Mental Health 20: 8699 (2010) Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/cbm.

757

Treatment engagement as a predictor of premature treatment termination and treatment outcome in a correctional outpatient sample

KLAUS H. DRIESCHNER1 AND JOOP VERSCHUUR2, 1Trajectum Hoeve Boschoord, The Netherlands; 2Forensic Outpatient Treatment Center De Tender/Oldenkotte, The Netherlands ABSTRACT Background It is widely assumed that the outcome of correctional treatment largely depends on the patients treatment engagement; however, evidence for this notion is scarce. Aim In this study, the Treatment Engagement Rating scale (TER) was used to investigate the relationship of a patients treatment engagement with premature treatment termination and treatment outcome, dened as reduction of the risk of reoffending. Method Employing a prospective design, this study addressed the relationship of the behavioural treatment engagement of correctional outpatients (N = 138) with treatment non-completion and treatment outcome, respectively. Results Treatment engagement as measured by the TER emerged as an accurate predictor of treatment non-completion (area under the curve = 0.76; odds ratio = 4.1) and was also signicantly correlated with treatment outcome ( = 0.41). The prediction was more accurate for sex offenders than for violent offenders, for expulsion from the treatment than for dropout, and when treatment engagement was assessed closer to the end of the treatment. Conclusions The study provides evidence that the outcome of correctional treatment depends to a substantial degree on the behavioural efforts that the patients make for the treatment. Periodic assessment of treatment engagement in correctional treatment facilities would be useful, and the TER seems to be a practical, reliable and valid instrument for this purpose. Copyright 2010 John Wiley & Sons, Ltd.

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010) DOI: 10.1002/cbm

Treatment engagement

87

Introduction Correctional treatment requires effort from the patient in order to be successful, regardless of the particular treatment approach, the criminogenic needs that have to be addressed and the treatment setting. Patients must attend sessions; disclose feelings, thoughts and facts; refrain from problem behaviour; try new behaviour; and accept the rules and restrictions imposed by the therapist or treatment facility. To the extent that making and sustaining this effort is aversive, it requires motivation from the patient. This motivation, in turn, depends on the patients problem recognition, perception of the treatment and therapists, outcome expectancy, distress, perceived legal pressure and experienced strain from the treatment. Most of these cognitive and emotional determinants of the patients motivation to make efforts for the treatment can be inuenced, either by specic interventions such as motivational interviewing and motivational enhancement therapy (McMurran, 2009), by a general motivational style (e.g. Kear-Colwell and Pollock, 1997) or by adherence to the responsivity principle, i.e. the adaptation of the treatment to the patients cultural background, learning style, cognitive capacity and so on (Kennedy, 2000). In a nutshell, the above outlined chain of hypothesised relations between treatment outcome, the patients efforts for the treatment (in the following termed treatment engagement), the motivation to make these efforts, the cognitive and affective internal determinants of this motivation, and the external inuences on these internal determinants constitute the integral model of treatment motivation and related concepts (Drieschner et al., 2004; see Figure 1). The central part of the model was largely supported in earlier studies with forensic outpatient samples (Drieschner and Boomsma, 2008a, 2008b). The present study is

External factors.
(e.g., legal pressure, social network)

Internal determinants of MET

Problem related

External factors

Treatment factors
(e.g., motivational interventions and style; adaptation to learning style, cognitive capacity, etc.)

LP DS CT ST

PR MET TE

Treatment outcome
Crimino genic needs Reoffending

Patient factors
(e.g., personality, age, psychopathology)

OE

Treatment efficacy

Treatment related

Figure 1: Integral model of treatment motivation and related concepts applied to correctional treatment. Note: LP = perceived Legal Pressure; PR = problem recognition; DS = distress; CT = perceived Costs of the Treatment; ST = perceived Suitability of the Treatment; OE = outcome expectancy; MET = Motivation to Engage in the Treatment; TE = treatment engagement

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010)

DOI: 10.1002/cbm

88

Drieschner and Verschuur

concerned with the right-hand part of the model, i.e. the relationship between treatment engagement and treatment outcome, from which the model derives its relevance. After all, there would be no point in motivating offenders to engage in treatment if the resulting behaviour had no impact on treatment outcome. The desired outcome of offender treatment is the prevention of reoffending. A methodological difculty of this outcome variable is that it can only be assessed years after treatment has ended. A more proximal outcome is the reduction of the patients criminogenic needs, i.e. dynamic risk factors that have an evidencebased association with recidivism. Another variable that has consistently been shown to be related to recidivism, and that has an even more proximal relation with treatment engagement, is premature treatment termination versus treatment completion (e.g. Hanson and Bussire, 1998; McMurran and Theodosi, 2007). Conceptually, this variable could be regarded as a component of treatment engagement. However, in several respects, it has more in common with treatment outcome than with other components of treatment engagement. Unlike treatment engagement, it cannot wax and wane during the treatment, and like treatment outcome, it becomes manifest by the end of the treatment. Moreover, there is a semantic overlap between treatment outcome and premature treatment termination because the fact that termination is regarded as premature implies that the treatment was not fully successful. These similarities with treatment outcome justify conceptualising premature termination versus treatment completion as a dependent variable and not as a component of treatment engagement. The predictive power of treatment engagement for premature termination and treatment outcome has been addressed by several studies with correctional samples. The association between treatment engagement and treatment outcome was found to be either moderate (Taft et al., 2003; Gondolf and Wernik, 2008), absent (Barbaree, 2005) or even negative (Seto and Barbaree, 1999). There is more evidence for the relationship between treatment engagement and premature termination (e.g. Wormith and Olver, 2002; Nunes and Cortoni, 2006). Studies into the impact of treatment engagement often suffer from an ambiguous conceptualisation and measurement of treatment engagement. Often measures are employed which entangle behavioural treatment engagement with cognitive variables, such as attitude towards the treatment, condence in the treatment process, the therapeutic alliance and (the perception of) treatment benet (e.g. Seto and Barbaree, 1999; Elbogen et al., 2006; Cunningham et al., 2009; Day et al., 2009). Other conceptually purer measures of treatment engagement have other limitations, such as globally dened items (Gondolf and Wernik, 2008) or applicability to only a particular type of treatment (Macgowan, 1997). A pure measure of treatment engagement with broad applicability is the Dutch Treatment Engagement Rating Scale (TER; Drieschner and Boomsma, 2008c).1 With the TER, eight components of treatment engagement are measured, which
1

A formally translated English version is available from the rst author.

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010) DOI: 10.1002/cbm

Treatment engagement

89

are relevant for most treatment approaches and most categories of patients in correctional treatment. The results of initial psychometric research were promising (for details, refer to the Method section). In the present study, the TER is applied to investigate the relationship of patients treatment engagement with premature treatment termination and treatment outcome, dened as reduction of the risk of offending. Method Settings and participants The study was conducted in a Dutch correctional outpatient treatment facility. Most patients received treatment as alternative to imprisonment or criminal prosecution, or in the rehabilitation phase following inpatient treatment. Sometimes, voluntary treatment was offered to individuals at risk of committing an offence. The treatment approach was based on cognitivebehavioural principles with the primary focus on prevention of reoffending. Treatments could include individual and group therapy, partner therapy, non-verbal therapy, social work and assertive community interventions. Most treatments lasted between 12 and 24 months with an intensity of one or two sessions a week, often followed by infrequent booster sessions during a period of 1 years. Ten per cent of the patients received a more intensive day treatment. The sample (N = 138) was a subsample of the one reported in Drieschner and Boomsma (2008c). To be included, patients had to be literate and in treatment for at least 15 weeks at the time of the TER rating with a minimal intensity of one session every 3 weeks. Note that the sample did not include patients who dropped out of the treatment during the rst 15 weeks of the treatment. Of the 214 patients who met the inclusion criteria, 13 refused informed consent, and in 16 cases the therapists did not provide TER scores. Of the remaining 185 patients, 11 did not give consent for use of their data in follow-up research, and two died during the treatment. In 34 cases, the clinical les did not contain the required information about treatment termination and treatment outcome or could not be obtained at all. For characteristics of the resulting sample (N = 138), see Table 1. Variables and measures The study design included one independent and two dependent variables. The independent variable was treatment engagement, dened as the patients behaviour which is desirable or necessary for the treatment to be effective and which is under the patients volitional control. Dependent variables were treatment completion versus non-completion and treatment outcome, dened as a reduction of criminogenic needs.

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010)

DOI: 10.1002/cbm

90

Drieschner and Verschuur

Table 1: Sample characteristics (N = 138) N Age Average number of years at intake (SD) Gender Male Ethnicity Both parents Dutch nationality At least one parent with Asian (including Turkish), African or South American nationality Other Type of offence (n = 137) Sexual offence Hands on, age of victim <16 years Hands on, age of victim 16 years Hands off Violent offence (no sexual component) Domestic Non-domestic (no economic motive) With economic motive Arson Economic offence without violence Miscellaneous (e.g. stalking, vandalism) No (known) offence Psychopathology DSM-IV axis I Psychotic Substance abuse Paraphilia Mood or anxiety disorder Impulse control disorder Other axis I disorder No axis I disorder DSM-IV axis II Personality disorder cluster B Personality disorder cluster C Borderline intellectual disability No axis II diagnosis available %

35 (11.6) 125 127 10 1 90.6 92.0 7.2 0.7

50 36 5 9 58 14 39 5 7 10 3 9

36.5 26.3 3.6 6.6 42.1 10.2 28.5 3.6 5.1 7.3 2.2 6.5

5 38 20 27 22 12 5 57 12 6 19

3.6 27.5 14.5 19.6 15.9 8.7 3.6 41.3 8.7 4.3 13.8

Note: DSM-IV = Diagnostic and Statistical Manual of Mental Disorder.

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010) DOI: 10.1002/cbm

Treatment engagement

91

TER The TER (Drieschner and Boomsma, 2008) is a Dutch therapist rating instrument for the behavioural efforts of patients in correctional outpatient treatment (see footnote 1). The instrument has 21 items with anchored 5-point response scales. The items address the following components of treatment engagement: Session Attendance (two items), Making Sacrices (three items), Openness (two items), Effort to Change Problem Behaviour (two items), Goal Directedness (two items), Efforts to Improve Socio-Economic Situation (ve items), Constructive Use of Therapy Sessions (three items), Dealing with the Content of Therapy Between Sessions (one item) and Global Rating of Treatment Engagement (one item). The TER generates nine component scores and a total score, which is the mean of the component scores. An earlier psychometric evaluation of the TER (Drieschner and Boomsma, 2008c) yielded the following results. The nine component scores represented a single dimension and had a high internal consistency (Cronbachs = 0.93; in the present sample = 0.94). The inter-rater reliability was satisfactory for the TER total score [intra-class correlation (ICC) = 0.76] and moderate for the component scores (ICC between 0.56 and 0.69). The validity of the TER total score was supported by correlations of 0.47 and 0.66 with the scores of the Motivation to Engage in the Treatment (MET) scale of the self-report Treatment Motivation Scales for Forensic Outpatient Treatment, and with a therapist rating of MET, respectively. The correlation between the TER scores and a latent factor representing the common variance of both MET instruments was 0.91. The TER takes between 5 and 10 minutes to complete. Treatment termination Type of treatment termination was retrospectively coded from the patient les using the following denitions. Dropout: the treatment was prematurely terminated by the patient without the therapists consent; expulsion: the treatment was prematurely terminated by the therapist due to rule violation by the patient other than no-show; termination in agreement but against advice: the therapist agreed with termination but advised continuation of the treatment; full treatment completion: treatment termination in mutual consent and in accordance with the therapists advice. Assigning the same value to dropout and expulsion resulted in a 3-point scale. For analyses that require a dichotomous dependent variable, full treatment completion and termination in agreement but against advice were also assigned the same score. To validate the treatment termination rating, six therapists still employed at the treatment centre were asked to code from memory the treatment termination and treatment outcome of 42 former patients using the denitions given above. In 14 cases, the therapists had no clear memory of the treatment. For the other 28 cases, a satisfactory level of agreement with the scores coded from the patient les was found, as indicated by a Kappa value of 0.73 for the dichotomised termination scores and an

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010)

DOI: 10.1002/cbm

92

Drieschner and Verschuur

ICC (single measure, two-way random) of 0.63 for the 3-point termination scale. Treatment outcome Based on information in the patient les in most cases the discharge letters four levels of treatment outcome were distinguished using the following denitions. (Almost) no change: no improvement in terms of risk reduction; unsatisfactory change: some improvement but no substantial reduction of risk; satisfactory change: improvement resulting in satisfactory risk reduction; highly successful: more than satisfactory improvement and reduction of risk. Analogue to the procedure for treatment termination, the validity of the coding was assessed with memorybased therapist ratings using the same denitions. The ICC (single measure, two-way random; N = 26) between the treatment outcome scores as coded from the patient les and as indicated by therapists was 0.74. Patient background variables Information about age, gender, type of offence and psychopathology (see Table 1) was extracted from the clinical les. Type of offence is based on the last conviction. When there were several convictions, the most serious offence was encoded. Almost all clinical les contained information about diagnoses on axes I and II of the Diagnostic and Statistical Manual of Mental Disorder (DSM-IV). In most cases, these were clinical diagnoses not based on a formal structured interview. Information about ethnicity (nationality of the patients parents) was provided by the patients at the time of the TER rating. Procedure The study was approved by a recognised Dutch medical ethical committee. Between December 2002 and December 2003, 17 therapists completed the TER for their patients, who had signed for informed consent on the basis of written information about the study. When several therapists were involved in the treatment of one patient, the mean of the TER scores was used in the analyses (for more details, see Drieschner and Boomsma, 2008a). At the time of the TER rating, the treatments were, on average, in their 57th week (SD = 30) and went on for another 79 weeks (SD = 61), including periods of aftercare with booster sessions. In August 2009, information about the treatment (i.e. duration, type of termination, outcome) and about the patient (i.e. DSM-IV diagnosis, type of offence) was extracted from the patients clinical les by the rst author, who at that time was blind for the patients TER scores. A few weeks later, the patients former therapists, who were blind for both the TER scores and the information extracted

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010) DOI: 10.1002/cbm

Treatment engagement

93

from the case les, recalled information concerning treatment termination and treatment outcome from their memory. Data analysis First, Spearman correlations were computed for the relation of treatment engagement with both treatment termination and treatment outcome. To explore the relevance of the various components of treatment engagement, this was conducted for the total score as well as the component scores of the TER. Second, receiver operating characteristics (ROC) analysis was employed to assess the predictive accuracy of TER total score for the dichotomised treatment termination. The ROC analysis produces an area under the curve (AUC) value, which equals the probability that a randomly chosen treatment completer gets a higher TER score than a non-completer. AUC values of 0.70 are widely interpreted as indicating moderate discrimination, and values above 0.75 as good discrimination. Third, a direct logistic regression analysis was conducted with the dichotomised treatment termination as dependent variable and the TER total score, diagnosis of substance abuse or addiction, presence of a cluster B personality disorder, age and gender as predictor variables. The purpose of this analysis was to investigate whether TER scores improve the prediction of premature treatment termination when recognised risk factors, such as substance abuse (e.g. Browne et al., 1998; Nunes and Cortoni, 2006), younger age (e.g. Chang and Saunders, 2002) and antisocial personality (e.g. Moore et al., 1999), are taken into account. Odds ratios were computed for variables that make a statistically signicant contribution to the prediction. Finally, the ROC analysis was repeated for separate sample fractions. First, separate analyses were conducted for patients with TER rating in the nal year of the treatment and patients who had more than a year of treatment to go. Predictions are likely to be more accurate when TER ratings are accomplished closer to the end of the treatment because there is less time for events and processes that affect the likelihood of premature termination. Second, separate AUC values were computed for sex offenders and violent offenders. Results Descriptive statistics for treatment termination and treatment outcome are presented in Table 2. Note that due to the inclusion criteria, the sample did not include patients who dropped out of treatment during the rst 15 weeks. This may explain the lower prevalence of treatment non-completion (33%) compared to the 45% found in other studies in correctional outpatient settings (McMurran and Theodosi, 2007). The Spearman correlation between the dependent variables, namely treatment outcome (four levels) and treatment termination (three levels), was 0.76.

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010)

DOI: 10.1002/cbm

94

Drieschner and Verschuur

Table 2: Descriptive statistics for treatment termination and treatment outcome n Treatment termination (N = 133)a Premature treatment termination Dropout Expulsion Treatment completion With agreement but against advice Full treatment completion Treatment outcome (N = 126)b Almost no change Unsatisfactory change Satisfactory change Highly successful change
a

44 30 14 89 47 42 14 44 53 15

33.1 22.6 10.5 66.9 35.3 31.6 11.1 34.9 42.1 11.9

No relevant information in clinical les of 5 patients. No relevant information in clinical les of 12 patients.

Table 3: Spearman correlations between the scores of the TER and treatment termination and treatment outcome, respectively TER component Treatment terminationa 0.25** c 0.28** 0.26** 0.49** 0.27** 0.36** c 0.44** 0.45** 0.44** 0.45** 0.49** e 0.45** g Treatment outcomeb 0.20* d 0.35** 0.26** 0.45** 0.28** 0.39** d 0.37** 0.38** 0.33** 0.41** 0.56** f 0.40** h

Session attendance Making sacrices Openness Effort to change problem behaviour Goal directedness Efforts for improving socio-economic situation Constructive use of therapy sessions Dealing with content of therapy between sessions Global rating of treatment engagement TER total Sex offenders Violent offenders

Note: TER = Treatment Engagement Rating scale. a N = 133; bN = 126; cN = 125; d N = 118; eN = 48; fN = 46; gN = 57; h N = 54. *p < 0.05 **p < 0.01

TER total scores were distributed normally with a mean of 3.33 and an SD of 0.75. The TER total score and all component scores had substantial and statistically signicant correlations with both treatment termination and treatment outcome (see Table 3). The component Effort to Change Problem Behaviour was

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010) DOI: 10.1002/cbm

Treatment engagement

95

the strongest predictor for both outcome indicators, even outperforming the TER total score. This is remarkable given the higher inter-rater reliability for the total score (ICC value of 0.76 versus 0.58; Drieschner and Boomsma, 2008c). The TER total score was also a strong predictor for treatment completion versus premature termination, as indicated by an AUC value of 0.76. Post hoc analyses revealed that the prediction was substantially more accurate for expulsion from the treatment (AUC = 0.87) than dropout (AUC = 0.71; see also Table 2). In the logistic regression model with the independent variables TER total score, age, gender, substance abuse and cluster B personality disorder, TER score was the only statistically signicant predictor of premature termination (see Table 4). A second model without the non-signicant predictors yielded an odds ratio for the TER scores of 4.09 (condence interval 2.247.45). This means that premature termination of the treatment was four times more likely to occur when the TER total score was 1 point (i.e. 1.33 standard deviations) lower.
Table 4: Direct logistic regression analysis of premature termination as a function of TER scores, age, gender, substance abuse and cluster B personality disorder Variables B Wald test z-ratio TER score Age Gender Substance abuse Personality disorder (cluster B) Constant 1.33 0.01 0.25 0.30 0.17 4.27 4.17 0.47 0.34 0.70 0.04 2.84 p <0.001 0.65 0.73 0.49 0.97 0.01 3.77 1.01 0.78 1.39 1.02 0.02 Odds ratio

Note: TER = Treatment Engagement Rating scale.

The predictive accuracy of the TER scores was slightly better for sex offenders (N = 50, AUC = 0.82) than for violent offender (N = 58, AUC = 0.72). There were no statistically signicant differences between both groups with respect to the proportion of premature termination, the prevalence of cluster B personality disorder and the number of weeks in treatment, either before or after the TER rating. However, the violent offender group was younger [mean age 31.6 versus 41.7 years, t(106) = 4.94, p < 0.001], had more individuals with diagnosis of substance abuse disorder or addiction [34.5% versus 12.0%, 2(1) = 7.43, p = 0.007] and had lower TER total scores [mean score 3.10 versus 3.42, t(106) = 2.11, p = 0.037]. As was expected, TER scores of patients in the nal year of the treatment (mean 25 weeks to go) predicted better than scores of patients who had longer than 1 year to go (mean 118 weeks). The Spearman correlations of the TER total

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010)

DOI: 10.1002/cbm

96

Drieschner and Verschuur

score with treatment termination and treatment outcome were 0.61 and 0.55, respectively, in the nal year sample fraction (41% of the sample) versus 0.31 and 0.15, respectively, in the long time to go sample fraction. Likewise, the AUC values for the differentiation between treatment completion and non-completion were 0.83 in the nal year group and 0.66 in the long time to go group. In the nal year group, premature termination was more prevalent [47% versus 25%, 2(1) = 6.20, p = 0.013], and TER scores were lower [t(123) = 2.21, p = 0.029]. The groups did not differ signicantly with respect to age, prevalence of either substance abuse or cluster B personality disorder, or number of weeks in treatment before the TER rating. Thus, the average treatment duration was shorter in the nal year group, possibly due to the higher prevalence of premature termination.

Discussion The study provides evidence that the outcome of correctional treatment depends, to a substantial degree, on the behavioural efforts that the patients make for the treatment. This nding is in accordance with the integral model of treatment motivation and related concepts (Drieschner et al., 2004). It is remarkable how accurately treatment outcome and premature termination were predicted by the patients treatment engagement, as measured by the TER, especially if it is taken into account that (1) the TER scores reected the patients behaviour during no more than 2 months, which was less than 10% of the average treatment; (2) treatment engagement can wax and wane during the treatment; and (3) the treatments continued, on average, more than 1 year after the TER rating, during which many events and processes could take place that affect the likelihood of premature termination. The dynamic character of the patients treatment engagement is emphasised by the nding that treatment engagement was a better predictor of treatment outcome and premature termination when assessed closer to the end of the treatment, independent of the duration of the treatment at the time of the TER rating. This underscores the importance of assessing treatment engagement repeatedly during the treatment. This would also enable investigation of the relationship between uctuations of treatment engagement and treatment outcome. The various TER components did not predict treatment outcome equally well. In accordance with the results of Taft et al. (2003), Session attendance had comparatively little predictive power. The strongest predictor was Efforts to change problem behaviour. This component consists of two items reecting efforts to change problem behaviour within therapy sessions (e.g. Efforts towards more adequate social behaviour, Efforts to control impulsive behaviour, Allowing oneself to be more vulnerable) and between therapy sessions (e.g. Avoiding risky situations, Improving communication with partner, Actively tackling problems), respectively. It

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010) DOI: 10.1002/cbm

Treatment engagement

97

is not surprising that the patients efforts to change problem behaviour predict treatment outcome. After all, change of problem behaviour is more or less equivalent to treatment outcome. Thus, the predictive power of the TER component Efforts to change problem behaviour may simply reect that the harder a patient tries, the more likely he or she is to succeed in changing problem behaviour, which intuitively makes sense. TER scores predicted expulsion from the treatment more accurately than dropout. A possible explanation for this nding is the substantive similarity between lack of treatment engagement, and reasons for expulsion. To illustrate, expulsion from the treatment could result from repeated verbal aggression towards therapists during therapy sessions, or from repeated alcohol abuse between sessions. The same behaviour would also result in low scores for the TER component Efforts to change problem behaviour. Put differently, expulsion from treatment can be a consequence of lack of treatment engagement. In accordance with ndings by Nunes and Cortoni (2006), treatment engagement was a slightly better predictor of treatment outcome for sex offenders than for violent offenders. However, given the small N, this difference must be interpreted with caution. The study has several limitations. First, it did not address the relationship between treatment engagement and reoffending, which undoubtedly is the most relevant outcome variable. However, given the strong association between treatment engagement and premature termination, which is a recognised strong predictor of reoffending (e.g. Wormith and Olver, 2002), TER scores are likely to predict reoffending to at least some degree. Second, the correlation between treatment engagement and treatment outcome does not imply that the latter is caused by the former. It could also be attributed to confounding variables that are correlated with both treatment engagement and treatment outcome. Some potential confounders, such as substance abuse, cluster B personality disorder and age, were ruled out in a logistic regression analysis, but others, such as social network and job situation, were not assessed. Further research into this issue is desirable because only if treatment engagement is a causal factor for treatment outcome can motivational interventions be expected to enhance treatment outcome. Third, retrospective coding from clinical les is certainly not the method of choice for the assessment of treatment outcome. Because in some cases the same therapist scored the TER and wrote the discharge letter, which was the primary source of information about treatment outcome in the patient les, the correlation between TER scores and treatment outcome might be inated by common method variance. Although the average period between TER rating and writing the discharge letter was more than 1 year, it is possible that truly independent measurements of treatment engagement and treatment outcome would have yielded a weaker correlation between both variables. Note, however, that this does not apply to the correlation between treatment engagement and premature termination. Finally, the reliability of the coding of

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010)

DOI: 10.1002/cbm

98

Drieschner and Verschuur

treatment termination and outcome from the patient les was not established by independent ratings. However, the fact that the coding was largely in accordance with the memory of the former therapists provided evidence that the information in the client les was extracted correctly. Despite these limitations, the present study provides an argument for structural periodic assessment of treatment engagement in correctional treatment facilities. The TER seems to be a practical, reliable and given the present results valid instrument for this purpose. References
Barbaree HE (2005) Psychopathy, treatment behavior, and recidivism. Journal of Interpersonal Violence 20: 11151131. Browne KD, Foreman L, Middleton D (1998) Predicting treatment drop-out in sex offenders. Child Abuse Review 7: 402419. Chang H, Saunders DG (2002) Predictors of attrition in two types of group programs for men who batter. Journal of Family Violence 17: 273292. Cunningham WS, Duffee DE, Huang Y, Steinke CM, Naccarato T (2009) On the meaning and measurement of engagement in youth residential treatment centers. Research on Social Work Practice 19: 6376. Day A, Howells K, Casey S, Ward T, Chambers JC, Birgden A (2009) Assessing treatment readiness in violent offenders. Journal of Interpersonal Violence 24: 618635. Drieschner KH, Boomsma A (2008a) The Treatment Motivation Scales for Forensic Outpatient Treatment (TMS-F): Construction and psychometric evaluation. Assessment 15: 224241. Drieschner KH, Boomsma A (2008b) Validation of the Treatment Motivation Scales for Forensic Outpatient Treatment (TMS-F). Assessment 15: 242255. Drieschner KH, Boomsma A (2008c) The Treatment Engagement Rating Scale (TER) for forensic outpatient treatment: description, psychometric properties, and norms. Psychology, Crime & Law 14: 299315. Drieschner KH, Lammers SMM, Van der Staak CPF (2004) Treatment motivation: An attempt for clarication of an ambiguous concept. Clinical Psychology Review 23: 11151137. Elbogen EB, Dorn RAV, Swanson JW, Swartz MS, Monahan J (2006) Treatment engagement and violence risk in mental disorders. British Journal of Psychiatry 189: 354360. Gondolf EW, Wernik H (2008) Clinician ratings of batterer treatment behaviors in predicting reassault. Journal of Interpersonal Violence 24: 17921815. Hanson RK, Bussire MT (1998) Predicting relapse: A meta-analysis of sexual offender recidivism studies. Journal of Consulting and Clinical Psychology 66: 348362. Kear-Colwell J, Pollock, P (1997) Motivation or confrontation: Which approach to the child sex offender? Criminal Justice and Behavior 24: 2033. Kennedy SM (2000) Treatment responsivity: Reducing recidivism by enhancing treatment effectiveness. Forum on Corrections Research 12: 1923. Macgowan MJ (1997) A measure of engagement for social group work: The Groupwork Engagement Measure (GEM). Journal of Social Service Research 23: 1737. McMurran M (2009) Motivational interviewing with offenders: A systematic review. Legal and Criminological Psychology 14: 83100. McMurran M, Theodosi E (2007) Is treatment non-completion associated with increased reconviction over no treatment? Psychology, Crime & Law 13: 333343. Moore DL, Bergman BA, Knox PL (1999) Predictors of sex offender treatment completion. Journal of Child Sexual Abuse 7: 7388.

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010) DOI: 10.1002/cbm

Treatment engagement

99

Nunes KL, Cortoni F (2006) Estimating risk of dropout and expulsion from correctional programs (Report No. 2006 R-177). Ottawa, Canada: Correctional Service Canada. Seto MC, Barbaree HE (1999) Psychopathology, treatment behavior, and sex offender recidivism. Journal of Interpersonal Violence 14: 12351248. Taft CT, Murphy CM, King DW, Musser PH, DeDeyn JM (2003) Process and treatment adherence factors in group cognitive-behavioral therapy for partner violent men. Journal of Consulting and Clinical Psychology 71: 812820. Wormith JS, Olver ME (2002) Offender treatment attrition and its relationship with risk, responsivity and recidivism. Criminal Justice and Behavior 29: 447471.

Address correspondence to: Klaus Drieschner, c/o Trajectum, Boylerstraat 4, 8387 XN Boschoord, The Netherlands. Email: kdrieschner@trajectum.info

Copyright 2010 John Wiley & Sons, Ltd

20: 8699 (2010)

DOI: 10.1002/cbm

Copyright of Criminal Behaviour & Mental Health is the property of John Wiley & Sons, Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.