In the past three decades, developed prison and probation services have paid large attention to risk assessment tools predicting the probability of reoffending. This not only resulted in a more accurate classification of offenders, but also in a more informed choice of effective intervention helping to reduce relapse in offending behaviour. In terms of reducing the risk of reoffending intervention programmes considering the principle of criminogenic risks, needs, and responsivity proved successful, while imprisonment on its own, where intervention methods were not applied, showed only limited effectiveness. For historical reasons, the Czech Prison Service underwent a different development, although its objectives were similar. It was not until the beginning of a new millennium when the Czech prison system together with a newly created probation service decided to seek new methods of assessing offenders based on criminogenic risks. This paper presents development and results of the first Czech tool used for assessment of offenders’ risks and needs, called SARPO (from the Czech abbreviation of Complex Analysis of Offenders’ Risk and Needs).
Offenders are imposed imprisonment sentences in order to prevent them from further offending, to rehabilitate them, and to protect the public. It means to protect the public not only in the time when the offender serves the prison sentence, but also after the offender is released. Developed prison and probation systems have lately turned their attention more and more to the latter aspect. They are aware of their share of responsibility for potential reoffending of the released offenders and they aim at mitigating the risk of reoffending, while assisting the released offender in reintegration into the community. For this purpose they have designed a large number of intervention methods.
In order to work with offenders it is necessary to have enough information about their attitudes towards themselves and others, what threats they pose to people that work with them and what chances of their rehabilitation there are. For this reason any intervention with the offender must be necessarily preceded by assessment mapping his/her personal features considering the situations in which the outputs will be used. These outputs may be different during imprisonment and after early release. Thus for a prison guard it will be important to know to what degree the offender is dangerous and whether he/she will present any risk of harm during the imprisonment both towards other people and to himself/herself. A therapist needs to assess any progress the offender has made in his/her own perception and any changes he/she has made in his/her attitudes. An early release board and the judge will be interested in knowing the likelihood of his/her reoffending after the release and of his/her law-abiding integration into the community. A probation officer will plan and organize the scheme of suitable intervention and make sure that the rehabilitation and reintegration of the released person is successful.
In the second half of the 20th century there was probably hardly a justice or prison system in the developed countries that would not be engaged in assessment and classification of offenders. A concise account of the development of assessment tools is given by Bonta [
According to Bonta, the first generation of assessment tools was clinical judgments, which were unstructured and based on experience of the assessor. Bonta believes this assessment to be inaccurate and unreliable leaving little space for consideration of other circumstances of offending. For this reason, in the 1980s, there came a second generation of tools using actuarial methods which started to emphasize so-called static factors, that is, those that could not be changed anymore: age, sex, start of the offending career, number of convictions, and other features typical of offending behaviour. This model showed higher predictive validity; nevertheless, it failed to match the criminogenic factors with intervention leading to rehabilitation of offenders. The 1990s brought about tools combining static and dynamic factors which built upon the premise that in order to be able to determine the risk of reoffending and to be able to choose a suitable intervention it is necessary to know both offender’s risks and criminogenic needs. While some of the tools of the third generation sum the items to arrive at numeric risk scores, other approaches incorporate evidence-based information about risk factors and expert interpretation of the seriousness, frequency, or duration of these risk factors—such as structured professional judgments (SPJ). Although the SPJ do not present higher predictive validity than actuarial tools, they help to concentrate on influencing the offender. Since the introduction of the SPJ it has not been sufficient to ask just whether and to what extent the offender is dangerous and also why it is so and what can be done in order to change that. Not later than at the turn of the millennium there appeared tools that take into account the length and intensity of the proposed intervention according to the seriousness of risks and needs: intervention and offenders must correspond; they must be responsive. This enables designing the intervention to individual offenders and at the same time to predict their future behaviour if certain conditions are met. Therefore, case management is considered by some authors an integral part of the fourth generation tools [
At present, all the four generations of tools are in use at once. Common praxis in the Anglo-Saxon countries shows that different approaches may well coexist with each other: the clinical rating scale Psychopathy Checklist-Revised (PCL-R) [
In the former Czechoslovakia, prediction of offending behaviour and reoffending started to be dealt with from the end of the 1960s. It was the Correctional Education Unit of The Penitentiary Research Institute that substantially contributed to this activity and that, based on the outputs of the assessment, researched also options for applying intervention methods as part of “treatment of prisoners” [
One of the first Czech prediction tools was Veverka’s Green Test which evaluated information relevant from a psychiatric point of view and his Blue Test focusing on data about offender’s criminal history and social background and identifying “risk” factors that make the offending career more likely [
A different approach when preparing his predictive critical sets of features was taken by Dobrotka [
For the time being, the last product in the Czech context is a working version of a five-item prediction instrument based on analysis of static data on the offending history of 100 persons drawn up by the Institute of Criminology and Social Prevention [
Compared to the development of predictive and assessment tools in the world, there were certain differences in the Czech environment. Although the tools intended to implement clinically relevant elements, in general they hardly ever regarded the possibility that an offender could change. Therefore, they concentrated mostly on static factors only. Some tools provided comprehensive prediction of reoffending, but had hardly any impact on prisoners’ treatment and rehabilitation. If taken in account, interventions were not aiming at mitigation of the risk of reoffending. The tools did not work with criminogenic factors in the sense of criminogenic risk and needs. If we were to classify the Czech tools within the above generations of tools used in the world, they would belong, with the exception of Dobrotka’s instrument, to the second generation of actuarial tools, even in cases where they included dynamic factors; the outputs of their dynamic part were not taken into account more than just by including them in numeric risk scores. And the most important one is that the tools never left the academic ground and were never employed in the judicial, correctional, and postpenitentiary routine.
At present, the Czech prison and justice systems use almost invariably psychodiagnosis and questionnaire methods, which either do not belong to assessment tools in terms of risk and needs assessment or which belong to the clinical generation. The emphasis is placed on individual approach; the collection of data from the prisoners is very large, while paying attention to the personality of the offender and obviously following the structure and function of judicial expert opinions. On the one hand, these are strengths that are traditionally and permanently inherent of the Czech environment. On the other hand, what may be seen as a weakness is that this information has not been used appropriately and its potential has been left unexplored. Thus, the acquired information has served the purpose of protection of the public in terms of reducing reoffending only to a limited degree. Therefore there is a need to have a completely new tool that would take into account the abovementioned risk and needs principles.
The assessment of offenders, as it has been perceived generally by the modern prison [
The objective of risk assessment is primarily to predict likelihood of offender’s reoffending and to adjust further intervention to the offender based on this prediction.
These variables, called
Research into criminal conduct [
Assessing criminogenic risk and needs helps—unlike other assessment methods—to define and address the management of prisoners as a whole: it helps to define who the target of the intervention is (in case of prisoners with high-risk of harm or risk of reoffending), what the intervention objectives are (mitigation of the risk of reoffending, that is, criminogenic needs, rather than, for example, clinical symptoms), what intervention strategy in terms of duration and intensity to take, while having in regard the motivation factors, the degree to which the offender is likely to be influenced and how he/she learns based on the responsivity principle (taking into account the current possibilities and abilities of the assessed person), and how the effectiveness of the intervention is measured.
The SARPO Project (from the Czech abbreviation Complex Analysis of Offenders’ Risk and Needs) is a joint project of the Czech Prison Service and the Czech Probation and Mediation Service which commenced as a part of the PHARE Project in 2003 by setting up a multidisciplinary team. At the beginning the team included employees of the Prison Service, Probation and Mediation Service, National Probation Service of England and Wales, the Czech Institute of Criminology and Social Prevention, City Centre of Social Services and Prevention, Association for Probation and Mediation in Justice, Municipal Court in Brno, State Prosecution Office in Prague 4, and the Police President Office of the Czech Republic. Once the PHARE Project ended in 2004, a team of prison and probation experts was established and strove to create its own assessment instrument which could be used by both Prison Service and Probation and Mediation Service.
Originally the SARPO Project was based on the Offender Management System assessment tool (OASys) which is currently used by the prison and probation service in England and Wales [
Basic parts of the SARPO 1 instrument and assessment mechanism.
SARPO uses static and dynamic risk factors to determine
The risk of reoffending reflects the likelihood of a future event of loss, damage, or harm. Here the risk of reoffending is mainly determined based on the offending history, and the likelihood of his/her further offending career is assessed by making a profile of the offender.
The risk itself includes
An important part of the tool is evaluation of offender’s
The basic structure of the tool consists of five parts. The first one evaluates information about the offender’s criminal history focusing on specific features of the offending. The second part of the tool analyses data relating to the circumstances, factors, and incentives relevant in the overall context of the offender’s criminal conduct, with a view to a potential object of harm. The third part is made of a set of dynamic risk factors (a total of 40) divided into six areas: housing, employment, finance situation, addictions, family and social relations, and attitudes and life-style. The fourth part contains a self-assessment form whose aim is to obtain feedback from the offender who himself/herself fills in this part. The last part is the final output where the results of the analysis of the offender’s risk and needs are interpreted, the offender’s motivation to address the difficult issues is assessed, and recommendation is made by the assessor, including proposed further treatment of the assessed person.
SARPO makes use of both objective and subjective methods of assessment. The objective method is represented by calculation of objectively determined risks on a scale from 0 to 10 points for each dynamic factor. Based on his/her justifiable expertise, the assessor is allowed to make a correction up to 2 points up or down. In addition to this, the calculated risks are compared with the self-assessment items made by the offender.
The pilot study designed to test the predictive validity of the SARPO instrument started in 2007 and was conducted by trained (all participating assessors underwent a three-day intensive training on using the instrument) assessors in 16 prisons and 19 probation centres across the Czech Republic. The original aim was to obtain a sample of 500 assessments in total: 350 by the Prison Service and 150 by the Probation and Mediation Service. The pilot study targeted two groups of respondents: prisoners prior to filing an application for an early release (Prison Service) and persons just after being released on parole. A total of 513 assessments were carried out: 379 prisoners and 134 persons on probation.
Evaluation of the pilot data was conducted in three modules over the next five years. In the first phase, basic information including quantitative, qualitative, and methodological evaluation was carried out and formulas for static and dynamic risk assessment were calculated. In the second phase, a certain correction of these formulas was suggested.
Finally, data on reincarceration of offenders from the Prison Service sample only were collected during spring 2012. Only those prisoners were involved who were actually granted conditional release from prison after the SARPO assessment and where the time period from their release lasted for no less than 3 years. Out of 379 tracked prisoners, only 282 were usable for the purpose of analysis of reoffending. The other prisoners were still in prison in 2012, or the time from their release was still shorter than 3 years. The analyzed dataset consists only of male offenders since the number of women included into SARPO study was minimal.
The indicator of recidivism used in this study is a new incarceration within 3 years after the release from prison. The recidivism rate is examined with relation to a set of static and dynamic risk factors. Static risk factors include age at the time of release, age of the first conviction, total number of convictions, crime intensity measured by Copas rate [
Dynamic risk factors—36 items with dichotomous scales coded 0 if the risk is not present and 1 if present—were divided into six groups which covered the following areas: accommodation, employment, finance, addictions, family/peers, and personality/attitudes. A separate risk scale for each group was constructed as a mean of respective items and the scales were subsequently linearly transformed to range from 0 to 10.
In addition, overall scales of static and dynamic risks were constructed. The dynamic risk scale was designed simply as a mean of six subscales (Cronbach’s alpha
Risk scales—descriptive statistics.
Mean | S.D. | Min/Max | Q1/Q3 | |
---|---|---|---|---|
Dynamic risks/needs subscales | ||||
Accommodation | 2.9 | 3.4 | 0/10 | 0/6.7 |
Employment | 4.5 | 3.4 | 0/10 | 1.7/6.7 |
Finance | 6.2 | 3.0 | 0/10 | 4.0/10.0 |
Addictions | 4.4 | 3.8 | 0/10 | 0/8.3 |
Family, peers | 4.7 | 3.0 | 0/10 | 2.0/7.0 |
Personality and attitudes | 4.3 | 2.6 | 0/10 | 2.0/6.0 |
Static risk scale | 46.3 | 18.5 | 2.6/83.2 | 32.6/59.0 |
Dynamic risk scale | 46.4 | 24.3 | 0/97.9 | 26.8/65.7 |
Overall risk scale | 46.4 | 18.4 | 2.0/84.4 | 32.3/60.9 |
Construction of all scales except for the static risk scale was developed independently from the reincarceration measure and therefore the strength of relationships to the dependent variable is not biased. We have adopted this approach because our sample size is limited and split-half cross validation is not possible. On the other hand, static risk scale was developed by fitting a logistic regression model with reincarceration as a dependent variable and therefore the strength of the relationship between the scale and reincarceration is likely to be overestimated and should be evaluated by further research.
Table
Descriptive statistics.
|
% | |
---|---|---|
Reincarceration | ||
Yes | 105 | 37.2 |
No | 177 | 62.8 |
Type of release | ||
Release | 131 | 46.5 |
Conditional release with supervision | 48 | 17.0 |
Conditional release | 103 | 36.5 |
Age when released | ||
19–29 | 99 | 35.1 |
30–39 | 109 | 38.7 |
40+ | 74 | 26.2 |
Education | ||
Less than primary | 20 | 7.1 |
Primary school | 109 | 38.7 |
Vocational school | 116 | 41.1 |
High school or higher | 36 | 12.8 |
No answer |
|
|
Marital status | ||
Married | 28 | 9.9 |
Single | 165 | 58.5 |
Divorced | 89 | 31.6 |
|
||
Total | 282 | 100.0 |
The associations between reincarceration and risk factors and scales are presented in Table
Risk factors and their association with reincarceration.
M+ (S.E.) | M− (S.E.) | Pearson | |
---|---|---|---|
Static risks | |||
First conviction (age) | 22.6 (0.56) | 19.2 (0.47) | −0.26 |
Number of convictions | 4.6 (0.25) | 6.2 (0.34) | 0.23 |
Crime intensity (Copas rate) | −1.8 (0.05) | −1.5 (0.05) | 0.26 |
Seriousness of offending | 0.24 (0.03) | 0.38 (0.05) | 0.14 |
Age | 35.3 (0.66) | 32.3 (0.85) | −0.16 |
Dynamic risks/needs | |||
Accommodation | 2.4 (0.24) | 3.8 (0.35) | 0.19 |
Employment | 3.7 (0.25) | 5.9 (0.31) | 0.32 |
Finance | 5.5 (0.22) | 7.5 (0.26) | 0.32 |
Addictions | 3.8 (0.27) | 5.6 (0.37) | 0.23 |
Family, peers | 4.0 (0.21) | 5.8 (0.30) | 0.28 |
Personality and attitudes | 3.7 (0.19) | 5.3 (0.23) | 0.31 |
Static risk scale | 41.0 (1.4) | 54.7 (1.6) | 0.36 |
Dynamic risk scale | 38.3 (1.7) | 56.2 (2.6) | 0.35 |
Overall risk scale | 38.7 (1.4) | 55.4 (1.6) | 0.42 |
All three scales—static, dynamic, and overall—seem to be fairly good predictors of recidivism and can well distinguish between low-risk and high-risk offenders as can be observed in Table
Proportion of reincarceration for quartiles of the risk scales.
Static risk | Dynamic risk | Overall risk | |
---|---|---|---|
Reincarceration (%) | Reincarceration (%) | Reincarceration (%) | |
1st quartile of the scale | 18.5 | 17.1 | 11.4 |
2nd quartile of the scale | 25.0 | 27.1 | 26.8 |
3rd quartile of the scale | 49.2 | 43.7 | 44.3 |
4th quartile of the scale | 63.1 | 60.0 | 66.2 |
ROC analysis for the overall risk scale.
The present results suggest that the instrument is well designed. It is not to expect that SARPO will accurately distinguish between those who fail from those who will commit no crime again. Therefore, we consider the current results as very good, especially as they come from a pilot study and are based on a limited sample size. The predictive ability of SARPO seems not to be much worse than analogous international tools (e.g., [
The results of the pilot study also clearly demonstrate the relevance of static and dynamic factors to estimate the risk of failure. On the other hand, it should be noted that this is only a pilot study based on a limited sample. Further research will be needed to evaluate the current construction of the static risk scale and to develop appropriate weights for subscales to create a more effective dynamic risk scale. It can be expected that the predictive ability of a modified tool will be even higher and will improve.
At this point we should mention why the Prison Service of the Czech Republic has decided to develop and introduce a new tool instead of implementing already established and proven international tools. Building a better tool for prediction of reoffending should not be the sole aim of risk prediction research, as it is just one link in the risk assessment—prediction-management. Risk assessment should lead to providing better treatment and continuity of care. We assume that the risk, needs, and responsivity principles provide a useful theoretical framework for interventions reducing risks of reoffending. The basic characteristics of Czech offenders do not seem to differ from populations of prisoners in other countries very much which gives support to the assumption that the majority of risk factors are similar across the developed countries. On the other hand, the tool SARPO was developed to refer to and recommend treatment solutions. Therefore a fundamental knowledge of the culture, cultural habits including drug use, kind of crime associated with the behavior of certain subcultures and ethnics, and also the legislative and socioeconomic system and social services is inevitable. Appropriate risk assessment can identify high-risk individuals in need of tailor-made management and available intervention. Using tools with dynamic risk predictors to assess risk can identify appropriate changeable treatment targets. Assessment tools with dynamic or changeable predictors, such as the HCR-20, OASys, or the LSI-R, can accomplish some of these tasks provided but are little connected with the available care options in the country.
More research involving prospective followups of large samples is clearly needed in the future. Since the tool SARPO has already been implemented into the practice of the Czech Prison Service, the number of assessments is rapidly increasing (over 9500 cases by the end of January 2014) which will soon allow us to conduct more elaborate analyses. In particular, there is a need to improve the predictive validity and to be more specific on dynamic risk scales.
Future directions and developments may lead SARPO towards assessing the difference between the risks identified by imprisonment and later on, either before applying for parole or on a regular basis. This means to develop, validate, and implement additional set of items which will be able to capture and interpret the relevant changes. This would open up completely new possibilities for evaluating the effectiveness of various interventions for various target groups of prisoners or to monitor the progress in therapy.
For the present, the Prison Service of the Czech Republic has a modern assessment tool, which has a number of advantages over simple static instruments or solely clinical tools.
The authors declare that there is no conflict of interests regarding the publication of this paper.