Abstract
The purpose of this study is to assess treatment change at both a group and individual level in a sample of 81 Dutch male patients who received mandated care for either violent (non-sexual) behavior or sexual violent behavior. Psychiatric nurses rated patients’ social skills, insight, hostility, physical violence with the BEST-Index every 6 months over the course of 2 years after patients were admitted to hospital. Mixed analysis of covariances and the reliable change index indicated that patients, irrespective of offense type, showed treatment change over time with exception of physical violence. This study shows that general treatment may be useful in the first 18 month for risk factors common to different types of offenses, but that specialized treatment is needed to establish further change.
Keywords
According to the World Health Organization (WHO, 2002, p. 4), violence refers to “the intentional use of physical force or power, threatened or actual, against oneself, another person, or against a group or community, that either results in or has a high likelihood of resulting in injury, death, psychological harm, maldevelopment or deprivation”. Most common types of violence are homicide, sexual aggression, intimate partner violence, and self-directed aggression. Each year, more than a million people lose their lives or suffer non-fatal injuries as a result of violence. Higher rates of violent behavior are found in institutionalized and secure settings (Huitema et al., 2021). This is not surprising because aggressive behavior is often the reason why patients are institutionalized or admitted to a secure setting in the first place. The cost of violence translates into billions of US dollars in annual health care expenditures worldwide, and billions more for national economies in terms of law enforcement and mandated care.
Establishing treatment change is consequently of particular importance for practitioners working in forensic settings as ineffective treatment has the potential to be detrimental for patients or elicit adverse outcomes for society (e.g., increased violence risk) (Mallion et al., 2020; Mews et al., 2017). Accordingly, (a) the overarching aim of mandated treatment is the reduction of risk-relevant behavior, which ultimately leads to a low risk for recidivism and (b) recidivism is one of the primary outcome measures in most studies and meta-analyses on treatment change of patients with offense histories.
General Risk Factors as Clinical Targets
Evaluations of treatment programs for persons with violent offenses in general show moderate reductions in reconviction rates after release (Henwood et al., 2015; Jolliffe & Farrington, 2004). For example, a review of psychological treatment for adults with violent offense histories reported a relative reduction of 23% in the proportion reconvicted for any offense (Henwood et al., 2015) after completing cognitive behavioral therapy. Focusing on treatment for sexual violence, meta-analyses indicate some level of treatment change (Olver et al., 2020; Schmucker & Lösel, 2015), although some studies report insufficient evidence (Dennis et al., 2012; Langstrom et al., 2013). For example, Schmucker and Lösel (2017) reported a significant reduction in recidivism rates in treated patients compared to patients receiving no treatment; the sexual recidivism rate was 10.1% versus 13.7%.
Treatment approaches that demonstrate the largest impact on reducing recidivism tend to adhere to the Risk-Need-Responsivity principles (RNR; Andrews & Bonta, 2017) with staff receiving regular supervision (Gannon et al., 2019; Hanson et al., 2009; Olver et al., 2011). The RNR framework underlines resources ought to vary in dose according to risk and should be aimed at reducing risk factors while being responsive to patient characteristics (e.g., intelligence, motivation). Amongst those general risk factors, four are particularly relevant to violent behavior. First, violent behavior in itself is a risk factor for future physical violence (Krupp et al., 2013). For example, offenders with an early age of arrest onset and more total arrest charges are more likely to commit serious criminal behavior (e.g., perpetrate rape, and murder) (Drury et al., 2017). Second, aggressive behaviors are believed to result from hostile cognitions and attributions (Eckhardt et al., 2004; Elmquist et al., 2016) that refer to a tendency toward negative judgment and dislike of others (e.g., denigration) (Birkley & Eckhardt, 2015; Norlander & Eckhardt, 2005). Hostility is not only in itself linked to violent behavior and recidivism (Garofalo et al., 2016; Kingston et al., 2009; Novo et al., 2012) but it is also associated with personality disorders (Nentjes et al., 2015) and substance use disorders (Smith Stover & Kiselica, 2015) which are prevalent in patients receiving mandated care. Third, a lack of social skills has been associated with various behavioral and developmental problems in children and adolescents, and with higher risk for (re-)offending in adults (Redondo et al., 2012; van der Put et al., 2012; van der Stouwe et al., 2016). Finally, a lack of insight (e.g., insight into disorder, problem behavior, precursors for criminal behavior) is a general risk factor for violent behavior (Lamb & Weinberger, 2017; Levenson & Macgowan, 2004). In treatment, physical violence, hostility, social skills and insight are generic risk factors, which are typically complemented by offense-specific risk factors such as antisocial orientation, poor self-regulation or sexual deviance in some persons with sexual offense histories.
Treatment Change: From a Group to an Individual Level
Typically, treatment change is measured on a group level. Patients receive treatment and recidivism rates are gathered upon release from a secure hospital, prison or community service, after which the data is analyzed as a group. Another way to assess treatment change is the approach to look at clinically significant change at an individual level. One way to do this is the method by Jacobson and Truax (1991) which involves calculating a reliable change index (RCI) which equals the difference between a patient’s pre- and post test scores, divided by the standard error of the differences. An RCI indicates whether an individual change score is statistically significantly greater than a difference that could have occurred due to random measurement error alone. This approach is used in psychological treatment in forensic psychiatry but is less common than group-level analyses (Chakhssi et al., 2010; Howard & van Doorn, 2018; Nunes et al., 2011; Wakeling et al., 2013). For example, Nunes et al. (2011) examined treatment change at both group and individual level with regard to risk, cognitive distortions, and general functioning in a sample of 313 adults with sexual offense histories. On a group level, almost all participants improved significantly (i.e., medium effect sizes), whereas the individual level analyses indicated more modest gains; approximately one third of the sample showed reliable change after treatment (Nunes et al., 2016). Howard and Van Doorn (2018) reported changes in antisocial attitudes over time that were significant at a group level whereas only a modest proportion of patients with offense histories achieved clinical significant change; less than four in ten were categorized as recovered. Reliable clinical change gives a greater depth of insight into treatment change than solely looking at group-level change.
Most research on treatment change for patients with offense histories focus on recidivism rates and thus examine change on a distal level. Less in known about change at a proximal level, thus change in risk relevant behavior during treatment while behavioral change is an important precondition for violence risk prediction. Moreover, assessment of risk relevant behavior is an ongoing process that forensic staff in general but psychiatric nurses in particular undertake on a daily basis (Walker et al., 2019). However, limited contemporary studies have reported on nurse observations on risk relevant behaviors in patients receiving mandated care. Chakhssi et al. (2010) examined treatment change in a sample of 74 patients with personality disorders who received forensic treatment. Psychiatric nurses assessed risk relevant behavior with the Behavior Status Index (BEST-Index; Reed et al., 2000) over the course of 20 months. Result showed significant improvement in social skills and insight, whereas change in physical violence and hostility was unchanged according to the reliable change index (Chakhssi et al., 2010; Jacobson & Truax, 1991). Similar results were reported in a study on the effects of a psycho-education program for a forensic population (n = 81) that showed an increase in adaptive social behavior as measured with the BEST-Index (Walker et al., 2012). In a single case study of a patient with psychopathic tendencies who received inpatient mandated care for committing a sexual offense, risk relevant behavior was assessed over the course of 4 years of treatment (Chakhssi et al., 2014). Reliable change analyses showed significant improvements in socials skills and insight, as measured with the BEST-index. After treatment, the patient was better at maintaining social relationships, had developed better social skills, and had developed more insight into his disorder. No change was reported with regard to physical violence and hostility, also measured with the BEST-index (Chakhssi et al., 2014). To our knowledge, with exception of Chakhssi et al. (2010), no research on the BEST-index is conducted on early treatment change in persons with sexual offense histories. Also, no studies seem to exist on differences between persons with sexual offense histories versus non-sexual violent offense histories.
Current Study
The current study is a prospective study to assess treatment change at both a group and individual level in a sample of 81 patients who received mandated care for either violent (non-sexual) behavior or sexual violent behavior. While most studies investigate treatment change at a distal level (i.e., recidivism), the current study investigates treatment change at a proximal level (i.e., change in clinical factors). We examined treatment change with regard to social skills, interpersonal hostility, insight and physical violence by means of the BEST-index. All patients received the same multi-modal general treatment approach. We expected improvement on a group and individual level in general risk factors after 2 years. On a group level, we predicted that there would be no differences between offense groups with regards to treatment change. Similarly, on an individual level, we expected that similar percentages of persons with sexual or violent offense histories would show improvement according to the reliable change index. Our hypotheses are based on the fact that we examined general risk factors that pertain to both sexual and non-sexual violence and the fact that the literature generally shows similar treatment change for both groups of offenders.
Method
Setting and Treatment
The study took place in a maximum security forensic hospital in The Netherlands. Patients receive mandatory treatment at the disposal of the state (in Dutch: “Ter Beschikking Stelling” [TBS]). The treatment order can be imposed to offenders if (1) the crime carries a maximum prison sentence of a minimum of 4 years; (2) the accountability for their crimes is judged to be (partly) diminished because of the presence of psychiatric diagnoses; and (3) if there is a continuous risk of recidivism without mandated treatment (Van Marle, 2002). TBS is aimed at reducing future violence risk by providing inpatient treatment and a stepwise rehabilitation (de Boer & Gerrits, 2007). According to Dutch criminal law (article 38b), the length of mandated treatment is indefinite but starts with 2 years. After 2 years, a criminal court of three judges can renew or terminate the TBS measure every 1 or 2 years based on the violence risk assessment and the proportionality of the duration of treatment related to the crime. The TBS measure can also be terminated provisionally during which the former patient can be recommitted if they recidivate or shows behavioral problems.
The most common forms of psychopathology in patients admitted to forensic psychiatric hospitals in the Netherlands are personality disorders, psychotic disorders, substance abuse disorders, paraphilias, and intellectual disabilities (Hildebrand & de Ruiter, 2004). On average, the length of stay in forensic hospitals in the Netherlands is 8 to 9 years (Brand & van Gemmert, 2009), during which the patient engages in a multi-modal treatment (e.g., psychological treatment, skills training, arts therapy, and vocational therapy). In the hospital where our study took place, patients live on a ward with 10 to 12 other patients who have similar treatment needs. The level of independence determines whether the ward atmosphere is either more group-focused or individual-focused (i.e., they can maintain a day structure by themselves). The treatment model of the hospital is eclectic with an emphasis on cognitive behavioral therapy. On the wards, they are offered a supportive milieu by psychiatric nurses.
Participants
The sample consisted of 81 male individuals who were convicted for either a violent offense (n = 41) or a sexual offense (n = 40). In order for our sample to be as homogeneous as possible, we did not include patients with intellectual disabilities: all participants had full scale IQ scores greater than 80 according to structured intelligence tests. Violent offending was defined as having committed a violent index offense (i.e., actual or grievous bodily harm, murder, manslaughter, assault) with no history of sexual offenses or sexual element to their offending. Sexual offending was defined as having committed one or more sexual offenses as defined by criminal law. In our sample, patients were either convicted for sexual abuse against children (n = 20) or against adults (n = 20). No patients were convicted for a non-contact sexual offense (e.g., use or distributing child sexual exploitation materials). No female patients participated in our study as there were no female patients admitted at the forensic hospital where the study was carried out. 1
For the entire sample (see Table 1), the average age at admission was 36.15 (SD = 10.44; range 20–70). Most prevalent DSM-IV axis I and II disorders were substance use disorders (63%), paraphilias (24.7%), PD Not Otherwise-Specified (PD NOS; 48.1%) and antisocial PD (19.8%). The average Psychopathy Checklist-Revised (Hare, 2003) score was 21.01 (SD = 6.82) and the average full scale IQ score was 95 (SD = 13.6).
Diagnoses per Offense Group.
Significant difference between groups at p < .01 with χ2 test.
Procedure
Approval for the study was obtained from the hospital’s executive board and the institutional review board. Two psychiatric nurses assessed every patient using the BEST-Index 6 months after admission and every 6 consecutive months up to 2 years, resulting in four time points in this study: 6 months after admission (T1), 12 months after admission (T2), 18 months after admission (T3), and 24 months after admission (T4). Assessments were performed by the primary nurse and one other nurse on the ward resulting in two nurses per patient who completed the assessment. Assessments were completed by 162 nurses in total. The interrater reliability coefficients for the BEST-Index scales and total score ranged from .76 to .99 at T1, .62 to .96 at T2, .73 to .96 at T3, and 64 to .99 at T4.
Measures
Diagnoses and demographic information were extracted from patients’ files. Diagnoses were based on semi-structured interviews made by certified psychiatrists and clinical psychologists. Scores in our dataset were part of the patient file and transferred to a data file. Because data collection included pre-existing scores from file, no information on inter-rater reliability was available.
Risk relevant behaviors
We assessed risk relevant behaviors with the BEST-index, which is a structured observational instrument used during inpatient treatment (Reed et al., 2000). The BEST-index consists of 70 items that are rated by psychiatric nurses over a 6-month period. The items are rated on a 5-point Likert-type scale (1 = worst case; 5 = optimal case). The study by Chakhssi et al. (2010) revealed a four-factor structure, including social skills, insight, interpersonal hostility, and physical violence, in additional to a total score (i.e., sum of items). The social skills factor (k = 23) refers to adaptive social behavior and social skills. The insight factor (k = 21) refers to insight that the patient has into the nature of his problems, into events that led up to their problems and into taking responsibility. The interpersonal hostility factor (k = 12) relates to a dominant or hostile stance in interpersonal interactions. The physical violence factor (k = 7) is related to whether the patient engages in physically violent behavior. We used the Dutch version of the BEST-index (Van Erven, 1999) which has shown adequate psychometric properties with interrater reliabilities ranging from .66 to .84 and good convergent validity of the BEST-Index with the HCR-20V2 (Chakhssi et al., 2010). In our study, the internal consistency of the factors ranged from Cronbach’s alphas of .77 to .95 at T1, .65 to .96 at T2, .72 to .97 at T3, and 77 to .97 at T4.
Analyses
We averaged the factor scores and total score between raters. Prior to our analyses, we examined potential differences between groups (i.e., sexual violence vs. non-sexual violence) in terms of clinical and demographic variables (i.e., age, PCL-R sore) using either Chi-square tests, independent samples t-test or Mann-Whitney U tests, depending on the distribution of our sample.
Our analytic approach consisted of two steps. First, we used a mixed analysis of covariances (ANCOVA) to examine differences in BEST-Index factors and total score between T1 and T4. Group (sexual violence vs. non-sexual violence) was used as between-subjects factor and the repeated BEST-Index assessments (i.e., four assessments over time) as within-subjects factor and psychotic disorders as a covariate. The BEST-index physical violence score was not normally distributed. For this factor, we used the Mann Whitney U test to examine between-group differences and the Friedman test to determine within-subject differences across the four time points. Second, we examined individual changes by using the reliable change index (RCI). The RCI provides a z-score where higher scores correspond with improvement and the threshold, for significant improvement (at p < .05) lies at a z-score ≥ 1.96. The RCI is calculated as:
X1 is a subject’s first BEST score and X2 the final BEST score (or reversed if lower scores mean better functioning). The Cronbach’s alpha is used as a reliability coefficient for the measure or subscale. Using categories suggested by Jacobson et al. (1999), we categorized individuals into one of these change categories following treatment:
(1) Deteriorated: an individual has demonstrated reliable change in the undesired direction (RCI < −1.96);
(2) Unchanged: an individual who demonstrated no reliable change (−1.96 < RCI < 1.96);
(3) Improved: an individual has demonstrated reliable change in the desired direction (RCI > 1.96).
Fisher’s exact tests for 3 × 2 associations were ran to examine group differences in reliable change. In both analytic steps, significant differences were investigated with a threshold adjusted for multiple comparisons (p = 0.05/5 = .01). All analyses were performed with the Statistical Package for the Social Sciences, version 27.
Results
Preliminary Analyses on Clinical Profiles
Group differences on axis I disorders were found in the prevalence of psychotic disorders (χ2(1) = 7.161, p = .007) which were more prevalent in the violent offense group. See Table 1 for the distribution of disorders across groups. The mean age at admission to the hospital was 38.38 for patients convicted for sexual violence and 33.98 for patients convicted for violent offenses. Group differences were non-significant (U = 974, z = 1.456, p = .145). The average PCL-R score was 19.77 and 21.99 for patients with convictions for sexual and violent offenses, respectively. Group differences were not significant (t(66) = 1.342, p = .184).
Change on Group Level
The mean scores and standard deviations for the BEST-index factors and total score across time score are presented in Table 2.
Mean Scores and Standard Deviations on the Different Time Points Between Groups.
Note. Higher scores indicate more favorable scores. The average length of time was 6.04 months (SD = 2.62) between T1 and T2, 6.35 (SD = 2.83) between T2 and T3, and 6.36 (SD = 3.05) between T3 and T4.
A mixed ANCOVA was run to determine the effects of treatment over time (i.e., four time points per BEST scale) between the two offender groups, when controlling for psychotic disorders. Analyses of the studentized residuals showed that there was normality, as assessed by the Shapiro-Wilk test, with exception of the physical violence factor score.
For the BEST-index social skills factor, the Mauchly’s test of sphericity indicated that the assumption of sphericity had been violated (p < .001, ε = .345). The main effect of time was statistically significant, F(2.666, 210.610) = 5.833, p < .001 (with Greenhouse-Geisser correction). The magnitude of the differences in means and effect size was very small (partial η2 = .069). The main effect of group was not significant, F(1) = 0.025, p = .875, nor the interaction effect (Group × Time), F(2.666, 210.610) = 0.957, p = .406 (with Greenhouse-Geisser correction).
For the BEST-index interpersonal hostility factor, the Mauchly’s test of sphericity indicated that the assumption of sphericity had been violated (p < .032, ε = .920). The main effect of time was statistically not significant, F(2.759, 217.950) = 2.238, p = .090 (with Greenhouse-Geisser correction). Both the main effect of group and the Group × Time effect were not significant (respectively, F(1) = .053, p = .818, and F(2.759, 217.950) = 1.383, p = .250).
For the BEST-index insight factor, the Mauchly’s test of sphericity indicated that the assumption of sphericity had been met (p = .113). The main effect of time was statistically significant, F(3, 237) = 22.677, p < .001. The magnitude of the differences in means and effect size was small (partial η2 = .223). Both the main effect of group and the Group × Time effect were not significant (respectively, F(1) = 1.074, p = .303 and F(3, 237) = 2.181, p = .091).
The BEST-Index physical violence factor was not normally distributed, therefore we used the Mann-Whitney U test for between group differences, indicating that there were no between-group differences for T1 (U = 898, z = .788, p = .430), T2 (U = 870, z = 0.493, p = .662), T3 (U = 956, z = 1.329, p = .184), and T4 (U = 999, z = 1.731, p = .083). A Friedman test was run to determine if there were within-subject differences in physical violence during the four time points. Physical violence decreased over time but the differences were not statistically different, χ2(3) = 1.660, p = .646.
For the BEST-index total score, the Mauchly’s test of sphericity indicated that the assumption of sphericity had been violated (p < .041, ε = .915). The main effect of time was statistically significant, F(2.746, 216.945) = 13.733, p < .001, partial η2 = .148 (with Greenhouse-Geisser correction). Both the main effect of group and the Group × Time effect were not significant (respectively, F(1) = 0.146, p = .146, and F(2.746, 216.945) = 0.874, p = .448; with Greenhouse-Geisser corrections).
Figure 1 provides a visual representation of the BEST scale scores and total score over time between the two offending groups. A post hoc paired samples t-test for the social factor showed no significant differences between 6 and 12 months (t(80) = −0.783, p = .436), a significant difference between 12 and 18 months (t(80) = −3.728, p < .001) but no significant change between 18 and 24 months (t(80) = −1.507, p = .136). For the insight factor, a post-hoc paired samples t-test showed no significant differences between 6 and 12 months (t(80) = −2.262, p = .026) and between 18 and 24 months (t(80) = −0.964, p = .338). There was, however, a significant difference between 12 and 18 months (t(80) = −4.437, p < .001). Similar findings were found for the BEST-Index total score. A post hoc paired samples t-test showed no significant differences between 6 and 12 months (t(80) = −1.972, p = .052), a significant difference between 12 and 18 months (t(80) = −4.438, p < .001) but no significant change between 18 and 24 months (t(80) = −0.733, p = .465).
Graphic overview of change over time.
Change at Individual Level
Table 3 presents the results of individual level change using the RCI, including the number and percentages of patients with violent and sexual offense histories that reliability improved, showed no reliable change, or reliably deteriorated. A Fisher exact test was used to examine whether there was an association between group and the RCI values per BEST-index factors and total score. There was no significant association between group and insight (p = .189), social skills (p = .134), interpersonal hostility (p = .039), and the total score (p = . 270). The Fisher exact tests was statistically significant for the physical violence factor (p = .003). A post hoc Fisher exact test in the 2 × 2 contingency table (deterioration vs. no change/improvement) revealed that a greater proportion of violent offenders reliably deteriorated on the physical violence factor during treatment, though the finding was non-significant (p = .043). No association was found between the groups and reliable improvement of the physical violence factor in the post hoc Fisher exact tests (improvement vs. deterioration/no change), p = .157.
Reliable Change for the BEST subscales and Total Score From T1 to T4 (n = 81).
Discussion
The main purpose of this prospective study was to assess early treatment change in a sample of patients with violent or sexual offense histories. We examined group and individual changes in risk relevant behaviors such as social skills, insight, physical violence, and interpersonal hostility as assessed with the BEST-index.
On a group level, statistical analyses showed that patients, irrespective of offense type, showed treatment change with regard to the BEST-Index social skills and insight factors. Closer examination showed that change occurred at 12 to 18 months of treatment. There were no significant changes in the first year which may be explained by difficulties in overcoming the coercive nature of their treatment. Change is dependent on the patients’ engagement in treatment, which has been characterized as low (Holdsworth et al., 2014; Sturgess et al., 2016). They may not be ready for treatment yet (Drieschner & Verschuur, 2010; McMurran & Ward, 2010), because they are perhaps not motivated to change due to the involuntary nature of their treatment, or lack insight to their problems (i.e., do not see need for treatment). There were no changes after 18 months of treatment which might indicate a ceiling effect. This suggests that there is a need for specialized treatment after 18 months. Alternatively, change in social skills and insight may also be dependent of the ability to practice social skills outside a forensic hospital setting (i.e., go on leave) which most patients may not have between 18 and 24 months (Barlow & Dickens, 2018; Sklenarova et al., 2020). There was no change over time for the BEST physical violence and hostility factors. Clinical significant change analyses using the reliable change index was used in order to explore individual change in risk relevant behavior. The results also showed that all BEST-Index factors with exception of physical violence did not differ between offense groups.
There may be several reasons for the lack of treatment change with regard to physical violence and hostility. First, forensic patients have limited choices regarding their treatment team and the fellow patients they spend time with. Because of the involuntary nature of their admission, patients may perceive their treatment as coercive. Coercive treatment can arouse strong negative feelings (Galon & Wineman, 2010; Hachtel et al., 2019). Furthermore, in our study, almost half of the patients were diagnosed with a personality disorder. Patients with personality traits are likely to have troubled interpersonal behaviors that complicate the adjustment to treatment (Fahlgren et al., 2020; Keulen-de Vos & de Vogel, 2021). Forced treatment may exacerbate this (Daffern et al., 2010). Second, physical violence and hostility are often the reason that these patients are admitted to mandated care in the first place. Staff members frequently encounter physical violence and hostility, either directed toward fellow-patients or directed toward staff. For example, recent research showed that in a study with 1,534 staff members who work with psychiatric patients, 65% had been a victim of at least one violent incident in the last 5 years (van Leeuwen & Harte, 2017). As a result, all patients’ behavior may be looked at with a “magnifying glass.” All behavioral infractions, mild, moderate or severe infractions, are monitored and reported. Relatedly, the fact that these patients have a criminal background, may lead some staff members to assume their behavior is more pathological than is actually the case. This may lead to a self-fulfilling prophecy (Keulen-de Vos et al., 2011). Third, we examined treatment change over a relatively short time period. In the Netherlands, the average forensic inpatient treatment is approximately 8 years. Because of the involuntary nature of their admission and the diagnostic profile of patients in mandated care, a treatment change within 2 years may be too short to observe changes in violent behavior.
Strengths, Limitations and Future Research
This study has a number of strengths that support the generalizability of our results. The sample was sufficiently large to allow for a meaningful analysis of the subgroups. Another strength was that it was a prospective study conducted in a forensic clinical setting and that forensic psychiatric nurses carried out the BEST-Index assessments as part of their clinical routine. The reliability of the measure was comparable to previous studies, supporting the ecological validity of our findings. Also, this study shows that the BEST-Index is a useful instrument to assess potential general risk-relevant behavior among forensic patients, including those with sexual offense histories. Finally, in addition to group-level analyses, we examined change during treatment in forensic patients on an individual level.
Some limitations of this study should be kept in mind when interpreting our findings. First, the lack of a no-treatment control group is problematic because it is not possible to determine whether observed change was caused by treatment or threats to internal validity such as the passage of time (Shadish et al., 2002). A relating factor is that the generalizability of our findings in limited to Dutch male inpatients. Future research should focus on different samples (e.g., outpatients, patients with intellectual disabilities, female patients, non-Dutch samples), and should focus on a longer observational period (>2 years). Second, we cannot rule out if staff habituation to patient behavior may have influenced the ratings. However, we had a large sample of different psychiatric nurses over time which decreases the likelihood for habituation effects. Third, we used an observational instrument to assess risk relevant behavior. Future research should incorporate various types of instruments to assess risk behavior during forensic treatment. Fourth, the RCI was calculated based on the first assessment which is after 6 months of treatment. Pre-treatment BEST scores were not available. Finally, it is unclear as to whether change in the earlier stage of treatment will correlate with successful outcome in the long run. Future research should examine the (causal) relationship between change in risk relevant behaviors and recidivism rates.
Conclusion
This study showed that general risk relevant behaviors such as social skills and insight can change in the first 2 years of forensic treatment in both inpatients with sexual and non-sexual violent offense histories. Observed physical violence and interpersonal hostility did not change over time which emphasizes the importance of specialized treatment of forensic patients. Our study also shows that group- and individual-level treatment change analyses provide complementary information; thus, both should be used in evaluation of forensic treatment. When representative norms are available, clinical significance provides a clear and established method for evaluating the progress of individual clients who receive mandated care.
Footnotes
Acknowledgements
We thank all patients and nurses for participating in this study and the directors of the hospital for their support and resources.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
