Longitudinal Change of Risk Factors During Treatment in High-Security Forensic Psychiatry in Flanders

Abstract

Empirical studies on the longitudinal change of risk factors during treatment of forensic psychiatric patients are scarce. Therefore, in the current study, changes in risk factors of 341 high-security forensic psychiatric patients in Flanders were investigated using latent growth curve analyses. Annual assessments of the Historical Clinical Future-Revised (HKT-R) at three time points were considered, starting at time of admission. A significant small improvement was found on the future scale and most future factors. However, the clinical scale did not change significantly and only one of the 14 clinical factors (i.e., problem insight) changed significantly over time. Although the forensic treatment seems to be a slow process, the sensitivity of the HKT-R to detect clinically relevant change is questioned, underscoring the importance of determining whether there are instruments capable of detecting small significant changes over time.

Keywords

forensic psychiatry risk factors latent growth curve analysis

Introduction

In the field of forensic psychiatry, a large body of research has studied risk and protective factors associated with recidivism (Bonta et al., 2014), as risk management is one of the primary goals of treatment. In this context, dynamic risk factors are especially important, because of their changeability over time (Douglas & Skeem, 2005). The degree of change in those dynamic risk factors is included in the examination and decision-making of practitioners about the risk of recidivism and the further treatment trajectory of an individual. However, only in the past decade, attention has shifted to the empirical assessment of longitudinal change of dynamic risk factors during treatment.

The most influential risk factors for recidivism in offenders were described by Andrews and Bonta (2010) as “the Central Eight”: four risk factors that are strongly associated with recidivism (i.e., history of antisocial behavior, an antisocial personality pattern, antisocial cognitions, and antisocial associations) and four risk factors that are moderately and indirectly associated with recidivism (i.e., a dysfunctional family situation, work and school problems, a lack of leisure activities, and substance abuse). This distinction was confirmed in mentally disordered offenders for violent recidivism (Bonta et al., 2014).

In recent decades, several risk assessment instruments have been developed to assess individual changes in empirically supported dynamic risk factors during treatment, for example, the Historical Clinical Risk Management-20 (HCR-20; Webster et al., 1997), the Violence Risk Scale (VRS; Wong & Gordon, 2006), and the Historical Clinical Future-Revised (Historische, Klinische, Toekomst-Revised; HKT-R; Bogaerts et al., 2018; Spreen et al., 2014). Since risk assessment and risk management are merged in practice, it is necessary to understand the changeable aspects of risk factors (Andrews & Bonta, 2010; Blanchard, 2013). However, progress can be made in the empirical assessment of the changeability of dynamic risk factors in inpatient forensic psychiatric patients since the literature is rather scarce.

Douglas and Belfrage (2001, as cited in Belfrage & Douglas, 2002; see also Webster et al., 2000) conducted the first study on changes of the HCR-20 over time in a forensic patient population. After approximately 6 months, significant moderate declines in the clinical and risk management scale were found in high-risk Canadian and Swedish forensic patients. Over the past two decades, multiple studies have been conducted in forensic psychiatric patients reporting significant small improvement on clinical scale scores during treatment (Belfrage & Douglas, 2002; Blanchard, 2013; Chakhssi et al., 2010; De Jonge et al., 2009; Hogan & Olver, 2016, 2019; Janković et al., 2021; Lewis et al., 2013; Mastromanno et al., 2018; Morrissey et al., 2014; Olsson et al., 2013; Olver et al., 2013; O’Shea & Dickens, 2015; Reed et al., 2000; Richter et al., 2018; Van der Linde et al., 2020; Webster et al., 2009). However, no significant change in dynamic factors was found in the study of Van der Veeken et al. (2018), using the Instrument for Forensic Treatment Evaluation (IFTE; Schuringa et al., 2014). The study was conducted in high security forensic patients at six time points with an interval of 4 to 6 months. Additionally, a Flemish study investigating a sample of patients residing in three forensic medium-security units found no significant difference in HCR-20 total scores between admission and discharge (Jeandarme et al., 2017). Similar non-significant findings were reported in the studies of Longdon et al. (2018) and Nitschke et al. (2020).

At the item level, De Jonge et al. (2009) found significant decreases on most clinical factors of the HKT-30 (Werkgroep Risicotaxatie Forensische Psychiatrie, 2002) after 2 years of treatment (problem insight, impulsivity, empathy, hostility, social skills, cooperation with treatment, responsibility for the offense, and coping skills), but the effect sizes were rather small. Other research confirmed the significant small decrease of several clinical factors during treatment, that is, lack of insight, negative attitudes, active symptoms of major mental illness, unresponsiveness to treatment, impulsivity, hostility, and egocentrism (Hildebrand & de Ruiter, 2012; Olsson et al., 2013; O’Shea & Dickens, 2015). Webster et al. (2000) described varying results, as significant decreases on all clinical factors of the HCR-20 were found in a high-security patient population in Canada, but only one clinical factor (i.e., negative attitudes) decreased significantly in Swedish high-security patients. Similarly, none of the clinical factors changed significantly in the study of Longdon et al. (2018).

Several risk assessment instruments, such as the HKT-30 and HCR-20, include risk factors related to the assessment of risk management in the future, after discharge. In the study of De Jonge et al. (2009), the HKT-30 future scale or risk management scale decreased significantly after about 2 years, as did all future factors. Several studies also reported significant small declines on the future scale (Blanchard, 2013; Douglas & Belfrage, 2001, as cited in Belfrage & Douglas, 2002; Hogan & Olver, 2016, 2019; Morrissey et al., 2014; Olsson et al., 2013). Although Belfrage and Douglas (2000) reported a decrease on the intra-institutional risk management scale after 1 year of follow-up, no change was found on the risk management scale for community release. Other authors also did not report significant changes on the future scale between assessments (Longdon et al., 2018; Nitschke et al., 2020; O’Shea & Dickens, 2015; Richter et al., 2018), while Mastromanno et al. (2018) found a significant increase after treatment.

It should be emphasized that various characteristics of the study design, the patient population or the type or duration of treatment or follow-up, can influence the findings and therefore explain differences between studies. For example, treatment duration seems relevant, as Van der Veeken et al. (2018) reported better resocialization skills in patients with a longer treatment history at the forensic psychiatric center (FPC) at first IFTE assessment. Janković et al. (2021) found a significant decrease in the clinical scale of the HKT-R during the first phase of treatment (period from juridical assessment until the first unguided leave), but a slower decrease during the second phase (period from the first unguided leave until unconditional release). This is in line with the finding of Richter et al. (2018), who found that only patients admitted for less than 2,112 days at baseline significantly decreased on the clinical scale. The greatest improvement in discharge plans of patients with longer prior treatment times seems to have occurred before the first assessment, so no significant improvements were reported afterwards. However, Van der Linde et al. (2020) found no differences in trajectories between different treatment duration groups.

Furthermore, psychopathology has been found to be associated with change (Janković et al., 2021). Patients diagnosed with substance use disorder or psychotic disorder showed a greater reduction in risk factors in the first phase of treatment, while patients with a cluster B personality disorder showed a greater decline in the second phase.

The same findings were seen in patients with psychotic disorder and cluster B personality disorder with or without a diagnosis of comorbid substance use. Moreover, the risk of recidivism could have an influence, as patients at higher risk at admission showed significant treatment progress, but an increase in problematic behavior was found in low risk patients (Van der Veeken et al., 2018). However, no different change trajectories were found between first-time offenders and recidivists, nor between different age groups (Van der Linde et al., 2020).

Lastly, according to Mastromanno et al. (2018), there might be differences in treatment between (1) patients who have been found not guilty by reason of mental impairment and (2) offenders who were found guilty of their crime and had served a prison sentence, but who were also diagnosed with a mental disorder and who have been referred to the mental health service. Possibly the focus is more on treating the mental disorder of these mentally ill prisoners than on rehabilitation and reintegration, which could explain why no decrease in the risk management scale for community release was found in their study.

It is important to address some issues raised in several studies that have not found significant decreases of dynamic risk factors during treatment. First, the sensitivity of the HCR-20 was questioned, as the three-point ordinal rating scale at item level may not be sensitive enough to detect change (Longdon et al., 2018; Morrissey et al., 2014; Nitschke et al., 2020; Olsson et al., 2013; Richter et al., 2018). The only study found that used a larger rating scale (i.e., IFTE) could not confirm this idea as no significant changes were discovered (Van der Veeken et al., 2018). The authors suggested that a 17-point rating scale might be too sensitive, detecting not only slight progress but also a decline, which could result in a mean change near zero. However, the prior treatment duration differed between patients at the start of IFTE assessment, which could also have influenced the findings. Second, most studies that found no significant change in dynamic risk factors during treatment applied a follow-up period of less than 2 years (Longdon et al., 2018; Nitschke et al., 2020, O’Shea & Dickens, 2015; Mastromanno et al., 2018). This study period may be too short to achieve significant improvement and longer intervals may be required to observe larger and positive changes.

In conclusion, since periodic risk assessments are an essential part of risk management, it is important to gain empirical understanding of the modifiable aspects of risk factors and to learn which factors would benefit from systematic increased attention during treatment (Andrews & Bonta, 2010; Blanchard, 2013). Most international studies found rather small but significant reductions in risk factors during treatment and after release, which were measured with different risk assessment instruments. However, research is still scarce, especially in Belgium, making this study even more relevant. The aim of the current study therefore contributes to deepening this line of research by assessing the longitudinal change of risk factors during treatment in high-security forensic patients in Flanders. The change of both the risk scales as the individual risk factors of the HKT-R were assessed. Based on prior research, we expected to find a small decrease in the clinical and future scales and risk factors of the HKT-R.

Method

Participants

Participants were 341 male patients (previously) residing in the two high security FPCs in Flanders. All patients were found not guilty by reason of insanity after having committed an offense and were subsequently admitted for mandatory treatment to diminish the risk of future offending. Mean age at admission was 40.7 years old (SD = 11.3, range = 19–77). Most of the patients (85.9%) were of Belgian nationality, followed by Moroccan (5.9%), Turkish (1.2%), Afghan (0.9%), and Dutch (0.9%) nationality. Based on the Diagnostic and Statistical Manual of Mental Disorders 5 (DSM-5; American Psychiatric Association, 2013), 38.1% of patients had a primary diagnosis of psychotic disorder, 27.9% Cluster B personality disorder, 8.2% other specified personality disorder, 8.2% paraphilic disorder, 4.7% substance abuse disorder, 4.7% conduct disorder, 4.1% neurobiological disorder, 0.9% Cluster A personality disorder, and 0.3% Cluster C personality disorder. The patients had committed a variety of index offenses. Among the cases with multiple index offenses, the most severe offense was selected: (attempted) homicide (23.5%), sexual offenses (23.2%), arson (4.4%), threats (5.9%), property crime (9.1%), drugs crime (1.5%), and other violent crimes (32.6%). The latter index offense included the offense of intentional assault and battery, as well as property offenses or extortion with violence, threats, or weapons.

Measure

The HKT-R is a Dutch risk assessment tool used to predict the likelihood of violent recidivism (Spreen et al., 2014) and is the revised successor of the HKT-30. It was created to support the structured professional judgment of the risk of violent recidivism in leave applications and TBS (“terbeschikkingstelling”; involuntary admission by order of the state for mentally disordered offenders) recommendations in the Netherlands. The goal was to develop an instrument to predict and identify characteristics of individuals for diagnosis and risk assessment purposes. Researchers and clinicians were consulted to include all elements deemed relevant.

The HKT-R consists of three scales, the historical, clinical, and future scale. The historical scale contains 12 historical factors related to the life history of the patient. These items are static risk factors and were therefore not investigated in the current study. The clinical scale consists of 14 clinical factors measuring the behavior of the patient in the previous 12 months (see Table 1). Lastly, the future scale includes seven future factors where the rater estimates the risks that emerge during leave modalities or after discharge from the FPC with further supervision (albeit at a lower level of supervision, i.e., supervised future scale), and after discharge from the FPC without further supervision (i.e., unsupervised future scale) (see Table 1). All risk factors are scored on a 5-point Likert scale ranging from 0 (“no risk”) to 4 (“high risk”). Therefore, a maximum of 56 points can be scored on the clinical scale and a maximum of 28 points on both the supervised and unsupervised future scale. After each HKT-R assessment, an individualized weighted structured professional judgment of the risk of recidivism is made by the clinicians for each patient. However, this final clinical judgment was not taken into account in this study. In the present study, the internal consistency of the three scales (clinical scale, supervised future scale, and unsupervised future scale) was considered acceptable at any time point, with Cronbach’s alphas ranging from .74 to .87 (Nunnally & Bernstein, 1994).

Table 1.

Means and Standard Deviations for HKT-R Clinical and Future Scales and Factors.

Variable	Time 1	Time 2	Time 3
Variable	M (SD)	M (SD)	M (SD)
Clinical scale	25.86 (9.30)	24.88 (9.63)	25.34 (9.63)
K01 Problem insight	2.86 (1.03)	2.68 (1.05)	2.66 (1.04)
K02 Psychotic symptoms	0.98 (1.39)	0.90 (1.27)	0.95 (1.33)
K03 Addiction	0.75 (1.25)	0.62 (1.15)	0.57 (1.11)
K04 Impulsivity	2.00 (1.37)	1.96 (1.33)	2.05 (1.35)
K05 Antisocial behavior	1.63 (1.19)	1.58 (1.12)	1.65 (1.14)
K06 Hostility	1.57 (1.17)	1.57 (1.13)	1.52 (1.10)
K07 Social skills	2.12 (1.00)	2.03 (0.97)	2.06 (0.94)
K08 Self-reliance	1.51 (1.19)	1.51 (1.24)	1.55 (1.27)
K09 Cooperation with treatment	2.12 (1.06)	2.01 (1.04)	2.07 (1.04)
K10 Responsibility for the offence	2.50 (1.05)	2.40 (1.07)	2.37 (1.14)
K11 Coping skills	2.53 (0.98)	2.50 (0.90)	2.54 (0.97)
K12 Violation of terms and agreements	2.22 (1.31)	2.19 (1.28)	2.20 (1.24)
K13 Labor skills	1.56 (1.41)	1.62 (1.39)	1.69 (1.37)
K14 Influence of risky network-members	1.45 (1.20)	1.32 (1.15)	1.31 (1.14)
Supervised future scale	12.99 (5.42)	11.45 (5.43)	10.15 (5.31)
T01 Agreement on future conditions	2.22 (1.10)	2.03 (1.13)	1.79 (1.19)
T02 Accommodation	1.58 (1.13)	1.22 (1.06)	0.88 (0.85)
T03 Financial situation	1.64 (1.13)	1.29 (1.14)	1.14 (1.09)
T04 Employment	1.82 (1.25)	1.60 (1.29)	1.42 (1.27)
T05 Daily activities	1.81 (1.21)	1.59 (1.22)	1.39 (1.22)
T06 Social network	1.75 (1.29)	1.55 (1.25)	1.54 (1.24)
T07 Stressing circumstances	2.20 (0.93)	2.15 (0.96)	2.00 (0.98)
Unsupervised future scale	22.08 (3.77)	21.04 (4.36)	19.97 (5.20)
T01 Agreement on future conditions	3.39 (0.63)	3.26 (0.68)	3.10 (0.82)
T02 Accommodation	3.22 (0.72)	3.06 (0.83)	2.80 (0.92)
T03 Financial situation	2.97 (1.01)	2.68 (1.14)	2.58 (1.13)
T04 Employment	3.19 (0.77)	3.09 (0.81)	2.93 (0.91)
T05 Daily activities	3.19 (0.69)	2.98 (0.90)	2.86 (0.92)
T06 Social network	2.64 (1.24)	2.56 (1.19)	2.55 (1.08)
T07 Stressing circumstances	3.53 (0.72)	3.44 (0.78)	3.34 (0.86)

Note. HKT-R = historical clinical future – revised.

Procedure

The original sample of the study consisted of all patients residing in the two high security FPCs in Flanders from the opening of the centers (November 17th, 2014, for FPC Ghent and August 7th, 2017, for FPC Antwerp) until July 3rd, 2020 (n = 633). Most patients were men (n = 618), as no women were admitted to FPC Ghent and only 10% of the admission capacity in FPC Antwerp can be allocated to female patients. All 15 female patients were excluded from the study because of the small number. Additionally, only patients with at least two HKT-R assessments were eligible for participation to be able to record change over time. Therefore, 200 patients were excluded as only one or no HKT-R assessment was done. This concerned patients who had only recently been admitted, patients who stayed in hospital for a short period of time, and patients for whom an HKT-R was not considered necessary because they had not committed a violent offense. Furthermore, 77 patients were excluded because of invalid HKT-R assessments that could affect the reliability of the measurements. An HKT-R assessment was considered invalid if more than two missing values were detected. This resulted in a final sample of 341 male patients.

The data were collected from the electronic patient files. The HKT-R was prospectively scored by psychologists during treatment, as part of routine outcome monitoring procedures. All psychologists followed a certified HKT-R training. According to good practice, a patient’s risk of recidivism should be assessed annually or when a transition to a lower security level is warranted. For this study, the first three HKT-R assessments were investigated. The first examination of the HKT-R (Time 1) for each patient was considered if it occurred within the first 2 years after admission, with a mean of 8.0 months (SD = 5.8, range = 0–24) after admission. On average, the second assessment (Time 2) took place 13.5 months (SD = 4.2, range = 5–35) after the first assessment and the third assessment (Time 3) followed 11.8 months (SD = 3.6, range = 1–25) after the second assessment. Therefore, there was a period of approximately 12 months between each time point. As previously mentioned, the HKT-R assessment was only considered valid if there were no more than two missing values. The scale scores were prorated when one or two missing values were present. In total, there were 312 valid assessments at Time 1, 341 valid assessments at Time 2, and 176 valid assessments at Time 3.

Statistical Analyses

Preliminary analyses were performed using IBM SPSS Statistics (version 25). All dependent variables were normally distributed, with skewness between −2 and +2 and kurtosis between −3 and +3 (Gunzler et al., 2021). Multivariate outliers were investigated and resulted in the removal of 109 values (0.80%). Missing data were processed using Full Information Maximum Likelihood (FIML) estimation to make optimal use of the data (Stoel & Garre, 2011).

Next, latent growth curve (LGC) analyses were calculated in Mplus version 7 (Muthén & Muthén, 1998–2012) to model changes in risk factors during treatment, from Time 1 until Time 3. The latent growth curve model (LGCM) is a method used to analyze longitudinal data and is suitable for investigating trajectories and changes of risk and protective factors over time. The LGCM compares lines of change across a range of individuals and determines the line of change of the overall model. LGCMs can be used to trace linear or curvilinear trajectories. Since the method uses structural equation modeling, models are adjusted for measurement errors. The mean baseline HKT-R scores (i.e., mean of intercept) and mean rate of change in HKT-R scores per assessment (i.e., mean of slope) were examined. Furthermore, the individual differences at baseline HKT-R scores (i.e., variance of intercept) and interindividual different rates of change per HKT-R assessment or growth trajectories (i.e., variance of slope) were described (Muthén & Muthén, 1998–2012). A total of 31 separate unconditional models were tested, in which the following dependent variables were separately included: the 14 clinical factors, the seven future factors at conditional release, the seven future factors at unconditional release, the clinical scale (i.e., the sum of the 14 clinical factors), the supervised future scale (i.e., the sum of the seven future factors at conditional release), and the unsupervised future scale (i.e., the sum of the seven future factors at unconditional release). A Bonferroni correction was applied for multiple testing (p < .0016). Model fit was assessed by global fit indices provided by Mplus: the Comparative Fit Index (CFI; >0.95 for reasonable fit) and the Standardized Root Mean Square Residual (SRMR; <0.08 for reasonable fit) (Hu & Bentler, 1999).

Finally, the RCI was calculated for the HKT-R clinical scale and factors, as well as the (un)supervised future scale and factors. The RCI determines if the individual change on each measure is statistically reliable and not due to measurement error. These analyses were performed because each HKT-R assessment was scored by a single rater and no conclusions could be made about inter-rater reliability. It was therefore unknown whether the scores were influenced by measurement error. The method of Jacobsen and Truax (1991) was adopted to calculate the RCI, in which the score at baseline was subtracted from the score at follow-up, then divided by the standard error of the difference. The standard error of the difference was calculated by √2(sd₁√1−r_xx)², where sd₁ is the standard deviation of baseline observations and r_xx is the reliability coefficient. The one-way random, single measure intraclass-correlation coefficient (ICC) was applied, which was assessed during the development of the HKT-R (Bogaerts et al., 2018; Spreen et al., 2014). The inter-rater reliability for the clinical domain was .85 (.67–.94), and for the future domain .42 (.09–.67). If the RCI was greater than 1.96, the null hypothesis was rejected that the difference was 0 with 95% confidence. The RCIs were calculated within tree assessment intervals, that is, between the first and second assessment, between the second and third assessment, and between the first and third assessment.

Results

Descriptives

Table 1 presents the descriptive statistics of the six risk scales and the 28 individual risk factors in the clinical and future domain, at the three time points.

Group-Level Change

Table 2 presents the mean and variance of the intercept and slope of the clinical scale and the 14 individual clinical factors. The LGC-model regarding the clinical scale and clinical factors fitted the data well. Since the chi-square is sensitive to sample size, only the CFI and SRMR were relied upon. The CFI and SRMR values were acceptable to good, meaning that only a small portion of the residual correlation between the observed variables could not be explained by the estimated model. The clinical scale did not decline significantly over time. Regarding the clinical factors, only one item declined significantly (K01, problem insight; slope of −0.13). Individual differences in slopes were found for the clinical scale and for K04 (impulsivity), K06 (hostility), K08 (self-reliance), K10 (responsibility for the offense), K11 (coping skills), K12 (violation of terms and agreements), K13 (labor skills), and K14 (influence of risky network-members), which signified significant different growth trajectories between individuals during treatment.

Table 2.

Parameter Estimates and Model-Fit Indices of Latent Growth Curve Model for the HKT-R Clinical Scale and Factors.

Variable	Parameter estimates				Model-fit indices
	Intercept		Slope		χ²			CFI	SRMR
	M	Δ	M	Δ	Value	df	p	CFI	SRMR
Clinical scale	25.85*	81.69*	−0.70	19.01*	2.66	1	.103	1.00	0.02
K01 Problem insight	2.85*	0.81*	−0.13*	0.10	3.66	1	.056	0.99	0.02
K02 Psychotic symptoms	0.99*	1.46*	−0.06	0.16	1.37	1	.242	1.00	0.01
K03 Addiction	0.72*	0.91*	−0.08	0.06	0.58	1	.447	1.00	0.01
K04 Impulsivity	2.01*	1.67*	−0.05	0.35*	0.16	1	.689	1.00	0.01
K05 Antisocial behavior	1.59*	1.14*	0.00	0.25	0.76	1	.382	1.00	0.01
K06 Hostility	1.54*	1.25*	0.01	0.33*	1.58	1	.209	1.00	0.02
K07 Social skills	2.13*	0.82*	−0.07	0.16	1.44	1	.231	1.00	0.02
K08 Self-reliance	1.53*	1.18*	−0.02	0.18*	0.44	2	.803	1.00	0.01
K09 Cooperation with treatment	2.12*	0.90*	−0.05	0.17	6.21	1	.013	0.98	0.03
K10 Responsibility for the offence	2.51*	0.94*	−0.09	0.17*	2.17	2	.338	1.00	0.02
K11 Coping skills	2.51*	0.82*	0.01	0.20*	1.03	1	.310	1.00	0.01
K12 Violation of terms and agreements	2.22*	1.29*	−0.02	0.32*	0.15	1	.704	1.00	0.01
K13 Labor skills	1.58*	1.74*	0.03	0.50*	1.07	2	.586	1.00	0.01
K14 Influence of risky network-members	1.44*	1.13*	−0.08	0.33*	1.62	2	.446	1.00	0.01

Note. HKT-R = historical clinical future - revised; CFI = comparative fit index; SRMR = standardized root mean square residual.

p ≤ .0016.

The parameter estimates and model-fit indices of the LGC models of the supervised and unsupervised future scales, as well as the 14 supervised and unsupervised future factors, are presented in Table 3. All models had a good fit to the data. A significant decline over time was found regarding the supervised future scale (slope of −1.29) and the unsupervised future scale (slope of −1.07). Looking at the future factors, significant negative slopes were present in the supervised future domain for T01 (agreement on future conditions; slope of −0.21), T02 (accommodation; slope of −0.34), T03 (financial situation; slope of −0.24), T04 (employment; slope of −0.18), and T05 (daily activities; slope of −0.17). In the unsupervised future domain, there was also a significant decline in T01 (slope of −0.13), T02 (slope of −0.19), T03 (slope of −0.23), T04 (slope of −0.13), and T05 (slope of −0.14). However, individual variations in between-subject trajectories could be found in all 16 models, except for T01 and T02 in the unsupervised domain. In the two latter cases, all individuals had similar rates of change per assessment.

Table 3.

Parameter Estimates and Model-Fit Indices of Latent Growth Curve Model for the HKT-R Supervised and Unsupervised Future Scale and Factors.

Variable	Parameter estimates				Model-fit indices
	Intercept		Slope		χ²			CFI	SRMR
	M	Δ	M	Δ	Value	df	p Value	CFI	SRMR
Supervised future scale	13.00*	26.28*	−1.29*	8.98*	3.09	2	.213	1.00	0.02
T01 Agreement on future conditions	2.24*	0.92*	−0.21*	0.41*	0.04	1	.836	1.00	0.00
T02 Accommodation	1.59*	1.27*	−0.34*	0.37*	0.76	2	.684	1.00	0.01
T03 Financial situation	1.62*	1.05*	−0.24*	0.39*	7.90	1	.005	0.96	0.04
T04 Employment	1.81*	1.20*	−0.18*	0.37*	1.34	1	.247	1.00	0.01
T05 Daily activities	1.80*	1.27*	−0.17*	0.39*	1.25	2	.537	1.00	0.01
T06 Social network	1.73*	1.37*	−0.13	0.42*	2.85	1	.091	0.99	0.02
T07 Stressing circumstances	2.20*	0.85*	−0.06	0.27*	0.59	1	.443	1.00	0.01
Unsupervised future scale	22.12*	11.56*	−1.07*	5.54*	0.00	1	.972	1.00	0.00
T01 Agreement on future conditions	3.39*	0.17*	−0.13*	0.08	0.01	1	.907	1.00	0.00
T02 Accommodation	3.23*	0.35*	−0.19*	0.12	0.55	1	.460	1.00	0.01
T03 Financial situation	2.98*	1.00*	−0.23*	0.35*	5.12	2	.77	0.99	0.05
T04 Employment	3.22*	0.36*	−0.13*	0.15*	0.60	1	.438	1.00	0.01
T05 Daily activities	3.17*	0.38*	−0.14*	0.17*	2.43	2	.297	1.00	0.02
T06 Social network	2.61*	1.19*	−0.02	0.24*	2.89	2	.235	1.00	0.03
T07 Stressing circumstances	3.53*	0.39*	−0.09	0.22*	0.08	1	.781	1.00	0.00

Note. HKT-R = historical clinical future - revised; CFI = comparative fit index; SRMR = standardized root mean square residual.

p ≤ .0016.

Individual Level Change

In Table 4 the RCI is presented, which represents the percentage of patients that reliably decreased, reliably increased, or showed no reliable change during treatment on the clinical and future domain. For each time period, the majority of patients did not show any reliable change on the clinical scale and clinical items. Similar results were found for the supervised and unsupervised future scale and items. The threshold values for each item to be considered a reliable change, were included in Table 4.

Table 4.

Reliable Change Index for the HKT-R Clinical and Future Scales and Factors (in %).

Variable	Time 1–Time 2					Time 2–Time 3					Time 1–Time 3
Variable	Threshold^a	N	Reliable decrease	No reliable change	Reliable increase	Threshold^a	N	Reliable decrease	No reliable change	Reliable increase	Threshold^a	N	Reliable decrease	No reliable change	Reliable increase
Clinical scale	10	301	8.31	86.05	5.65	10	176	1.70	92.05	6.25	10	137	16.06	72.26	11.68
K01 Problem insight	2	302	7.28	90.40	2.32	2	177	2.26	93.22	4.52	2	138	7.25	89.86	2.90
K02 Psychotic symptoms	4	295	0.34	98.98	0.68	4	172	0.00	100.00	0.00	4	133	1.50	98.50	0.00
K03 Addiction	2	297	11.78	82.49	5.72	2	175	6.29	89.14	4.57	2	135	10.37	84.44	5.19
K04 Impulsivity	3	302	1.32	97.02	1.66	3	177	2.26	96.61	1.13	3	138	0.00	92.75	7.25
K05 Antisocial behavior	3	299	0.67	98.33	1.00	3	176	0.57	91.73	1.70	3	136	2.21	95.59	2.21
K06 Hostility	3	300	0.67	98.00	1.33	3	176	1.14	98.30	0.57	3	137	2.19	93.43	4.38
K07 Social skills	3	300	0.33	99.67	0.00	2	176	4.55	90.34	5.11	3	137	0.73	98.54	0.73
K08 Self-reliance	3	300	0.67	98.67	0.67	3	175	0.00	100.00	0.00	3	136	0.74	97.79	1.47
K09 Cooperation with treatment	3	300	1.33	98.67	0.00	3	176	0.00	98.30	1.70	3	137	2.19	96.35	1.46
K10 Responsibility for the offence	2	286	4.55	94.06	1.40	2	165	1.82	96.97	1.21	2	127	7.87	88.19	3.94
K11 Coping skills	2	297	2.02	93.94	4.04	2	177	5.08	88.14	6.78	2	136	8.09	81.62	10.29
K12 Violation of terms and agreements	3	302	2.98	93.71	3.31	3	177	3.39	93.79	2.82	3	138	7.97	89.86	2.17
K13 Labor skills	2	284	8.80	82.39	8.80	2	175	6.29	85.71	8.00	2	133	13.53	72.18	14.29
K14 Influence of risky network- members	3	299	3.34	95.32	1.34	3	175	1.14	97.71	1.14	3	135	5.93	90.37	3.70
Supervised future scale	12	302	2.98	97.02	0.00	12	177	2.26	97.74	0.00	12	138	8.70	88.41	2.90
T01 Agreement on future conditions	3	300	3.33	94.00	2.67	3	176	3.41	95.45	1.14	3	136	11.03	84.56	4.41
T02 Accommodation	3	299	2.34	97.66	0.00	2	176	7.95	90.34	1.70	3	138	10.14	89.13	0.72
T03 Financial situation	3	302	4.30	95.36	0.33	3	177	0.56	97.74	1.69	3	138	6.52	90.58	2.90
T04 Employment	3	300	4.67	93.67	1.67	3	177	2.82	95.48	1.69	3	138	7.97	89.13	2.90
T05 Daily activities	3	297	4.04	94.95	1.01	3	176	2.84	96.02	1.14	3	135	9.63	88.89	1.48
T06 Social network	3	298	5.37	92.62	2.01	3	174	2.87	94.83	2.30	3	135	5.19	90.37	4.44
T07 Stressing circumstances	3	301	0.00	100.00	0.00	3	176	0.00	98.86	1.14	3	137	0.00	99.27	0.73
Unsupervised future scale	8	301	5.65	92.69	1.66	10	176	3.41	96.59	0.00	8	137	11.68	87.59	0.73
T01 Agreement on future conditions	2	298	3.36	94.97	1.34	2	174	4.60	94.83	0.57	2	134	8.96	89.55	1.49
T02 Accommodation	2	299	6.02	92.64	1.34	2	174	8.62	90.23	1.15	2	135	12.59	85.93	1.48
T03 Financial situation	3	301	3.99	96.01	0.00	3	177	3.39	94.35	2.26	3	138	7.97	89.86	2.17
T04 Employment	2	297	3.70	94.61	1.68	2	173	4.62	94.80	0.58	2	134	8.21	88.81	2.99
T05 Daily activities	2	294	7.82	91.16	1.02	2	173	4.05	94.80	1.16	2	132	8.33	90.15	1.52
T06 Social network	3	297	3.03	95.29	1.68	3	172	0.00	98.84	1.16	3	133	0.00	96.99	3.01
T07 Stressing circumstances	2	299	5.02	92.31	2.68	2	176	3.98	93.18	2.84	2	137	9.49	83.94	6.57

Note. HKT-R = historical clinical future – revised.

The threshold values for each item to be considered a reliable change.

Discussion

The aim of this study was to empirically measure longitudinal change of the risk scales and risk factors of the HKT-R in 341 male patients after 2 years of treatment in the two FPCs in Flanders. Based on group-level analyses, the clinical scale did not change significantly, contrary to findings of previous literature that reported a significant small decline (Belfrage & Douglas, 2002; De Jonge et al., 2009; Hogan & Olver, 2016, 2019; Janković et al., 2021; Jeandarme et al., 2017; Mastromanno et al., 2018; Morrissey et al., 2014; O’Shea & Dickens, 2015; Olsson et al., 2013; Richter et al., 2018; Van der Linde et al., 2020). Furthermore, a decline was observed in only one clinical factor. There was a significant improvement of problem insight of the patients, similarly to several previous studies (De Jonge et al., 2009; O’Shea & Dickens, 2015; Olsson et al., 2013). However, no significant change was detected in most clinical factors (Longdon et al., 2018). These findings are similar to the results of the individual-level analyses, as most patients did not change reliably on the clinical scale and all clinical factors across all time periods.

Regarding the future domain, previous studies empirically assessing the evaluation of the future scale, reported varying results, from no significant changes (Belfrage & Douglas, 2002; O’Shea & Dickens, 2015) to significant increases (Mastromanno et al., 2018) and reductions (Blanchard, 2013; De Jonge et al., 2009; Hogan & Olver, 2016, 2019; Morrissey et al., 2014; Olsson et al., 2013) over time. Since the current study made use of the future factors of the HKT-R, which is related to the future factors of the HKT-30, the findings are similar to the study by De Jonge et al. (2009). Indeed, a significant decline was found in the supervised and unsupervised future scale. Five future factors improved significantly in both conditions ((un)supervised). However, the significant reductions were rather small. These findings were again in line with the fact that most patients did not change reliably on the future scales and future factors across all time periods.

Implications

As discussed, we found no significant change after 2 years of treatment on the clinical scale and clinical factors, with the exception of one significantly decreasing clinical factor in the group-level analyses. Thus, the following question arises: Does the treatment have any effect on risk factors in the clinical domain? Before trying to answer this question, more information on the treatment program is warranted. Treatment programs in the FPCs are based on cognitive behavioral therapy, which is one of the most empirically supported interventions in the forensic context (Bonta & Andrews, 2017). Different therapeutic modules are developed and offered to patients based on their needs or dynamic risk factors. For example, as the module aggression regulation therapy targets impulsivity, hostility, and antisocial behavior, it is offered to patients with a Cluster B personality disorder or patients with similar needs. The different modules consist of verbal group therapies as well as non-verbal group therapies such as drama therapy, music therapy, art therapy, and psychomotor therapy. Each module is evaluated every 6 months by the multidisciplinary team.

Should a patient follow the full offer of therapeutic modules that is assigned to them, in the correct intensity and duration, it can be estimated that the therapeutic offer should be sufficient to treat the risk factors and eventually lower the risk of recidivism. However, several factors might hinder such a positive evolution. First, the required intensity of therapy (i.e., an average of four to five modules each week) is structurally not always feasible, since there are many patients in an FPC but only a relatively small capacity of treatment services. In turn, this structurally leads to a larger offer of group therapy, while some patients could benefit from an individual therapy program in addition to their group therapy. Thus, the patients included in the current study may not yet be following all the modules designated for them. Third, the motivation of the forensic patient population should be considered. There is sometimes a lack of internal motivation, which impacts the therapy progress and can even cause a patient to drop out of the module, even though therapy is mandatory. Lastly, it is possible that the skills learned by patients in therapy are not exercised sufficiently in everyday life on the ward, as is theoretically expected. Structural and in-depth discussions between nursing staff and therapists are planned on a weekly basis. However, due to the high workload, these evaluations tend to focus more on practical affairs or problems, resulting in less of a focus on the content of the therapy sessions. Therefore, the nursing staff might not be informed in sufficient detail in how they can support patients in applying the skills learned during therapy in stressful circumstances or in daily life. Additional effort in informing psychiatric nursing staff about the content of the therapy modules and the relevant skills for each individual patient, seems important to significantly improve on these risk factors, such as impulsivity, hostility, and coping skills. In turn, this could have a positive impact on other risk factors, such as the violation of conditions and cooperation with treatment.

Furthermore, it should be noted that some factors that did not change significantly, such as psychotic symptoms and self-reliance, as well as hostility and impulsivity may be specific to psychotic patients. However, since not all patients in our study were diagnosed with psychotic disorders, it is possible that changes may become apparent when the analyses are performed only on patients with psychotic symptoms (Janković et al., 2021). The same conclusion can be drawn for patients who exhibit antisocial behavior. Therefore, future research should include psychopathology as predictor to empirically verify this potential explanation.

Additionally, the follow-up period of the current study seems rather short because previous research has indicated that the reduction of risk factors is a rather slow process (Hildebrand & de Ruiter, 2012). Therefore, a two-year period might be too short to make significant progress in diminishing the risk of recidivism in a high-security patient population. Since most of these high-risk patients are characterized by multiple problems, they need to successfully follow multiple therapy modules to treat the different risk factors. Furthermore, it often takes some time before patients commit to their therapeutic trajectory. Establishing trust between therapists and patients is an essential prerequisite of a secure therapeutic relationship and a necessity to achieve treatment progress. Trust between patients is also important in group therapy, since patients are expected to share personal information with each other and be open to feedback from fellow patients. However, establishing a sufficient level of trust is not an easy task, as many patients have developed cognitive distortions during their prior prison sentence. Therefore, it would be an added value if patients were to move more quickly from a prison to an FPC after the internment sentence. In December 2019, 537 of those found not guilty by reason of mental impairment in Belgium still resided in prisons (Venken et al. v. Belgium, 2021) due to a lack of admission beds in forensic psychiatric facilities. Research has shown that a prolonged stay in prison can cause certain symptoms (e.g., paranoia, inability to engage in relationships, hampered decision-making, spatial disorientation), that are summarized by the term post-incarceration syndrome. These symptoms relate to the development of an antisocial attitude, a decrease in sensorimotor activities and an increase in passivity in several areas. When patients are admitted to an FPC after a long prison term, it takes a lot of time to adapt to the new clinical environment, which demands self-reflection and treatment cooperation which is not evident (Liem & Kunst, 2013). Thus, the effects of long-term incarceration may have influenced the results, especially for patients who have not been transferred from prison that long.

Moreover, the HKT-R may not be specific enough to evaluate short-term changes and to evaluate the treatment progress in sufficient detail. This is in line with the earlier reported doubts about the sensitivity of the HCR-20 to detect change (Longdon et al., 2018; Morrissey et al., 2014; Nitschke et al., 2020; Olsson et al., 2013; Richter et al., 2018). The five-point ordinal rating scale at item level of the HKT-R might not be sensitive enough to detect changes in the first months or years of treatment. This issue was already raised by Schuringa et al. (2014; 2018), who proposed to use a larger rating scale, since they are more likely to detect important minimal change. However, no evidence is currently available to support this suggestion.

Lastly, the fact that most patients did not change reliably on risk factors during treatment based on the individual-level analyses, warrants some further discussion. After approximately 2 years of treatment (i.e., between the first and third assessment), a two point change on the 5-point scale was considered a reliable change for 10 items, a three point change for 17 items and a four point change for one item. Similar thresholds exist after approximately 1 year of treatment (i.e., between the first and second assessment and between the second and third assessment). Rather large levels of change are therefore necessary for the change to be considered reliable, partly because of the poor to moderate ICC of most of the clinical and future items. However, this amount of change might not be feasible for high-security patients after 1 or 2 years of treatment. In comparison, the study by Van der Linde et al. (2020) in Dutch forensic psychiatric patients reported an average decrease of 11.73 points on the clinical scale of the HKT-R between the time of judicial psychiatric assessment and unconditional release. The average treatment time was 5.9 years, while in the current study a decrease of 10 points should be achieved in 1 year to be considered a reliable change. Additionally, the study of Van der Linde et al. (2020) consisted of only patients who were successfully treated, leading to unconditional release, while our study included patients independent of their treatment progress.

Strengths and Limitations

This is the first study in Flanders to empirically assess the longitudinal change of dynamic risk factors during treatment in a high-security forensic setting. Additionally, all high-security/risk patients residing in the FPCs from the opening of the centers until July 3rd 2020 were considered. To receive more reliable results, only patients with valid HKT-R scores were actually included in the study. This led to the total sample of 341 patients, which is a relatively high number of participants and similar to prior studies concerning the evolution of risk factors during treatment. Another strength of the current study is that we did not solely assess the evolution in the different risk scales, but also focused on the different risk factors individually. Even though the risk scales are very valuable to evaluate the risk of recidivism as a whole, it does not inform treatment practitioners on which specific risk factors need more attention. However, this knowledge is essential to make the necessary adjustments in treatment guidelines and ultimately reduce the risk of recidivism.

Apart from the strengths, this study also has some limitations. First, since the HKT-R assessments were scored prospectively by one rater as part of risk assessment during treatment, no inter-rater reliability could be calculated. The findings can therefore be questioned since it is unknown whether the observed changes are the result of measurement error. Second, the variance of the slope was significant for most clinical and future factors, but these interindividual differences were not assessed any further in the current study. Future research should therefore include predictors to the LGC model to try and explain these interindividual differences. As mentioned earlier, psychopathology might play a role (Janković et al., 2021), but the risk at admission could also be relevant (van der Veeken et al., 2018). Also, the duration of the prior prison sentence differed between the participants (in years, M = 4.9, SD = 5.6, range = 0.02–30.6), but was not controlled for in the analyses. As mentioned previously, a prior prison sentence can negatively affect the mental well-being and therefore hinder treatment progress. Lastly, the change in risk factors was only measured based on three time points, covering a period of 2 years. Since the average treatment time of completed treatments in the FPC is 2.8 years (Jeandarme et al., 2020), for most patients, the complete duration of the treatment was not considered. Future studies could include more time-points to set out trajectories from admission until release into society or in an institution with a lower security level.

Conclusion

Since reducing risk of recidivism is the main goal of forensic treatment, the empirical assessment of longitudinal change of risk factors during treatment is an important part of treatment evaluation. Applying LGC analysis to a high-security forensic population in Flanders, significant decreases were found among the HKT-R supervised and unsupervised future scale after 2 years of treatment, as well as on five future factors individually (i.e., agreement on future conditions, accommodation, financial situation, employment, and daily activities). However, no significant decline was found in the clinical scale, as well as most of the clinical factors. Only difficulties regarding problem insight ameliorated significantly. Additionally, most patients did not change reliably on all clinical and future factors. Therefore, a period of 2 years seems insufficient to decrease the HKT-R risk factors significantly and reliably. In addition to addressing difficulties associated with the treatment process in forensic psychiatry, the sensitivity of the HKT-R to detect clinically relevant change should be challenged.

Footnotes

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.

ORCID iD

Sophie Verschueren

References

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorder (5th ed.). https://doi.org/10.1176/appi.books.9780890425596

Andrews

D. A.

Bonta

(2010). The psychology of criminal conduct (5th ed.). Anderson Publishing Co.

Belfrage

Douglas

K. S.

(2002). Treatment effects on forensic psychiatric patients measured with the HCR-20 violence risk assessment scheme. International Journal of Forensic Mental Health, 1(1), 25–36. https://doi.org/10.1080/14999013.2002.10471158

Blanchard

A. J. E.

(2013). Dynamic risk factors in violence risk assessment: A multiple time-point evaluation of the HCR-20 and START [Unpublished master’s dissertation]. Simon Fraser University.

Bogaerts

Spreen

ter Horst

Gerlsma

(2018). Predictive validity of the HKT-R risk assessment tool: Two and 5-Year violent recidivism in a nationwide sample of Dutch forensic psychiatric patients. International Journal of Offender Therapy and Comparative Criminology, 62(8), 2259–2270. https://doi.org/10.1177/0306624X17717128

Bonta

Andrews

D. A.

(2017). The psychology of criminal conduct (6th ed.). Routledge.

Bonta

Blais

Wilson

H. A.

(2014). A theoretically informed meta-analysis of the risk for general and violent recidivism for mentally disordered offenders. Aggression and Violent Behavior, 19(3), 278–287. https://doi.org/10.1016/j.avb.2014.04.014

Chakhssi

de Ruiter

Bernstein

(2010). Change during forensic treatment in psychopathic versus nonpsychopathic offenders. The Journal of Forensic Psychiatry & Psychology, 21(5), 660–682. https://doi.org/10.1080/14789949.2010.483283

De Jonge

Nijman

H. L. I.

Lammers

S. M. M

. (2009). Gedragsverandering tijdens tbsbehandeling: Een multicenteronderzoek [Behavioural changes during forensic psychiatric treatment: A multicenter study]. Tijdschrift voor Psychiatrie, 51, 205–215. https://www.tijdschriftvoorpsychiatrie.nl/media/5/articles_2764pdf.pdf

10.

Douglas

K. S.

Skeem

J. L.

(2005). Violence risk assessment: Getting specific about being dynamic. Psychology, Public Policy, and Law, 11(3), 347. https://doi.org/10.1037/1076-8971.11.3.347

11.

Gunzler

D. D.

Perzynski

A. T.

Carle

A. C.

(2021). Structural equation modeling for health and medicine. Chapman and Hall/CRC. https://doi.org/10.1201/9780203701133

12.

Hildebrand

de Ruiter

(2012). Psychopathic traits and change on indicators of dynamic risk factors during inpatient forensic psychiatric treatment. International Journal of Law and Psychiatry, 35(4), 276–288. https://doi.org/10.1016/j.ijlp.2012.04.001

13.

Hogan

N. R.

Olver

M. E.

(2016). Assessing risk for aggression in forensic psychiatric inpatients: An examination of five measures. Law and Human Behavior, 40(3), 233. https://doi.org/10.1037/lhb0000179

14.

Hogan

N. R.

Olver

M. E.

(2019). Static and dynamic assessment of violence risk among discharged forensic patients. Criminal Justice and Behavior, 46(7), 923–938. https://doi.org/10.1177/0093854819846526

15.

L. T.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6(1), 1–55. https://doi.org/10.1080/10705519909540118

16.

Jacobsen

N. S.

Truax

(1991). Clinical significance: A statistical approach to defining meaningful change in psychotherapy research. Journal of Consulting and Clinical Psychology, 59(1), 12–19. https://doi.org/10.1037/0022-006X.59.1.12

17.

Janković

Van Boxtel

Masthoff

De Caluwé

Bogaerts

(2021). The long-term changes in dynamic risk and protective factors over time in a nationwide sample of Dutch forensic psychiatric patients. Frontiers in Psychiatry, 12, 737–846. https://doi.org/10.3389/fpsyt.2021.737846

18.

Jeandarme

Heesch

Boel

Dekkers

Goktas

Verbeke

(2020). High security geïnterneerden: Wie zijn zij? Waar komen ze vandaan? Waar gaan zij (niet) naartoe? [High security internees: Who are they? Where do they come from? Where are they (not) going?]. Panopticon, 41(5), 448–466.

19.

Jeandarme

Wittouck

Vander Laenen

Pouls

Oei

T. I.

Bogaerts

(2017). Critical incidents and the subsequent judicial reaction during medium security forensic psychiatric treatment in Flanders, Belgium. International Journal of Law and Psychiatry, 51, 54–61. https://doi.org/10.1016/j.ijlp.2016.12.001

20.

Lewis

Olver

M. E.

Wong

S. C.

(2013). The Violence Risk Scale: Predictive validity and linking changes in risk with violent recidivism in a sample of high-risk offenders with psychopathic traits. Assessment, 20(2), 150–164. https://doi.org/10.1177/1073191112441242

21.

Liem

Kunst

(2013). Is there a recognizable post-incarceration syndrome among released “lifers”? International Journal of Law and Psychiatry, 36(3–4), 333–337. https://doi-org.tilburguniversity.idm.oclc.org/10.1016/j.ijlp.2013.04.012

22.

Longdon

Edworthy

Resnick

Byrne

Clarke

Cheung

Khalifa

(2018). Patient characteristics and outcome measurement in a low secure forensic hospital. Criminal Behaviour and Mental Health, 28(3), 255–269. https://doi.org/10.1002/cbm.2062

23.

Mastromanno

Brookstein

D. M.

Ogloff

J. R.

Campbell

Chu

C. M.

Daffern

(2018). Assessing change in dynamic risk factors in forensic psychiatric inpatients: Relationship with psychopathy and recidivism. The Journal of Forensic Psychiatry & Psychology, 29(2), 323–336. https://doi.org/10.1080/14789949.2017.1377277

24.

Morrissey

Beeley

Milton

(2014). Longitudinal HCR-20 scores in a high-secure psychiatric hospital. Criminal Behaviour and Mental Health, 24(3), 169–180. https://doi.org/10.1002/cbm.1893

25.

Muthén

L. K.

Muthén

B. O.

(1998–2012). Mplus user’s guide (7th ed.). Author.

26.

Nitschke

Sünkel

Mokros

(2020). Forensic prevention outpatient clinic: A treatment evaluation study. The Journal of Forensic Psychiatry & Psychology, 31(5), 714–730. https://doi.org/10.1080/14789949.2020.1785527

27.

Nunnally

Bernstein

(1994). Psychometric theory. McGraw-Hill Higher, Inc.

28.

Olsson

Strand

Kristiansen

Sjöling

Asplund

(2013). Decreased risk for violence in patients admitted to forensic care, measured with the HCR-20. Archives of Psychiatric Nursing, 27(4), 191–197. https://doi.org/10.1016/j.apnu.2013.03.004

29.

Olver

M. E.

Lewis

Wong

S. C. P.

(2013). Risk reduction treatment of high-risk psychopathic offenders: The relationship of psychopathy and treatment change to violent recidivism. Personality Disorders, 4(2), 160–167. https://doi.org/10.1037/a0029769

30.

O’Shea

L. E.

Dickens

G. L.

(2015). Predictive validity of the Short-Term Assessment of Risk and Treatability (START) for aggression and self-harm in a secure mental health service: Gender differences. International Journal of Forensic Mental Health, 14(2), 132–146. https://doi.org/10.1080/14999013.2015.1033112

31.

Reed

Woods

Robinson

(2000). Behavioural status index (BESTIndex): A ‘life skills’ assessment for selecting and monitoring therapy in mental health care. Psychometric Press.

32.

Richter

M. S.

O’Reilly

O’Sullivan

O’Flynn

Corvin

Donohoe

Coyle

Davoren

Higgins

Byrne

Nutley

Nulty

Sharma

O’Connell

Kennedy

H. G.

(2018). Prospective observational cohort study of ‘treatment as usual’ over four years for patients with schizophrenia in a national forensic hospital. BMC Psychiatry, 18(1), 1–15. https://doi.org/10.1186/s12888-018-1862-0

33.

Schuringa

Heininga

V. E.

Spreen

Bogaerts

(2018). Concurrent and predictive validity of the instrument for forensic treatment evaluation: From risk assessment to routine, multidisciplinary treatment evaluation. International Journal of Offender Therapy and Comparative Criminology, 62(5), 1281–1299.

34.

Schuringa

Spreen

Bogaerts

(2014). Inter-rater and test-retest reliability, internal consistency, and factorial structure of the instrument for forensic treatment evaluation. Journal of Forensic Psychology Practice, 14(2), 127–144. https://doi.org/10.1080/15228932.2014.897536.

35.

Spreen

Brand

Ter Horst

Bogaerts

(2014). Handleiding en methodologische verantwoording HKT-R, Historisch, Klinische en Toekomstige-Revisie [Manual and methodological justification HKT-R, Historic, Clinical and Future-Revision]. Dr. van Mesdag kliniek.

36.

Stoel

R. D.

Garre

F. G.

(2011). Growth curve analysis using multilevel regression and structural equation modeling. In Hox

Roberts

J. K.

(Eds.), Handbook of advanced multilevel analysis (pp. 97–114). Routledge.

37.

Van der Linde

Bogaerts

Garofalo

Blaauw

De Caluwé

Billen

Spreen

. (2020). Trajectories of dynamic risk factors during forensic treatment: Growth trajectory of clinical risk factors in a sample of Dutch forensic patients. International Journal of Offender Therapy and Comparative Criminology, 64(15), 1491–1513. https://doi.org/10.1177/0306624X20909219

38.

Van der Veeken

F. C. A.

Lucieer

Bogaerts

. (2018). Forensic psychiatric treatment evaluation: The clinical evaluation of treatment progress with repeated forensic routine outcome monitoring measures. International Journal of Law and Psychiatry, 57, 9–16. https://doi.org/10.1016/j.ijlp.2017.12.002

39.

Venken et al. v. Belgium, ECHR 6 april 2021, nrs. 46130/14, 76251/14, 42969/16, 45455/17 and 236/19, ECLI:CE:ECHR:2021:0406JUD004613014, paragraph. 107.

40.

Webster

C. D.

Douglas

K. S.

Belfrage

Link

B. G.

(2000). Capturing change: An approach to managing violence and improving mental health. In Hodgins

(Ed.), Violence among the mentally ill: Effective treatments and management strategies (pp. 119–144). Springer.

41.

Webster

C. D.

Douglas

K. S.

Hart

S. D.

(1997). HCR-20: Assessing risk for violence (Version 2). Simon Fraser University.

42.

Webster

C. D.

Martin

M.-L.

Brink

J. H.

Nicholls

T. L.

Desmarais

S. L.

(2009). Short-term assessment risk and treatability (START). Forensic Psychiatric Services Commission.

43.

Werkgroep Risicotaxatie Forensische Psychiatrie [Taskforce Risk Assessment Forensic Psychiatry]. (2002). Handleiding HKT-30, versie 2002: Risicotaxatie in de forensische psychiatrie [Manual HKT-30, version 2002: Risk assessment in forensic psychiatry]. Ministerie van Justitie, Dienst Justitiële Inrichtingen.

44.

Wong

S. C.

Gordon

(2006). The validity and reliability of the Violence Risk Scale: A treatment-friendly violence risk assessment tool. Psychology, Public Policy, and Law, 12(3), 279. https://doi.org/10.1037/1076-8971.12.3.279