International Journal of Law and Psychiatry 32 (2009) 400–407

Contents lists available at ScienceDirect

International Journal of Law and Psychiatry

The precarious practice of forensic psychiatric risk assessments

Thomas Nilsson a,⁎, Christian Munthe b, Christina Gustavson c, Anders Forsman a, Henrik Anckarsäter a,c

a Forensic Psychiatry, Institute of Neuroscience and Physiology, the Sahlgrenska Academy, University of Gothenburg, Swedenb Department of Philosophy, Linguistics and Theory of Science, University of Gothenburg, Swedenc Forensic Psychiatry, Institute of Clinical Sciences, Malmö, Lund University, Sweden

⁎ Corresponding author. Forensic Psychiatry, LillhagspSweden. Tel.: +46 31 343 73 90.

E-mail address: thomas.nilsson@neuro.gu.se (T. Nils

0160-2527/$ – see front matter © 2009 Elsevier Ltd. Aldoi:10.1016/j.ijlp.2009.09.010

a b s t r a c t

a r t i c l e i n f o

Keywords:

Risk assessmentForensic psychiatryMentally disordered offendersViolenceEthical dilemmas

The development of forensic psychiatric risk assessments is discussed from a clinical point of view using theexample of Sweden. A central task in forensic psychiatry has traditionally been to identify dangerous,mentally disordered subjects considered to be prone to commit violent acts. Over time, “dangerousness” hasbeen reworded into “risk”. Nevertheless, such assessments have generally been based on the psychiatricfactors characterising the individual patient, while group interaction, situational factors, or social andcultural circumstances, such as the availability of alcohol and drugs, have been largely overlooked. That riskassessments have a focused on people with a diagnosis of “mental disorder” and been used as grounds forcoercive measures and integrity violations has somehow been accepted as a matter of course in the publicand political debate. Even the basic question whether offenders with a mental disorder are really more proneto criminal recidivism than other offenders seems to have been treated light-handedly and dealt with merelyby epidemiological comparisons between groups of persons with broad ranges of psychosocial vulnerabilityand the general population. Legal texts, instructions and guidelines from the authorities in charge are oftenvague and general, while actors in the judicial system seem to put their trust in psychiatric opinions. Theexchange of professional opinions, general public expectations, and judicial decision processes poses a hugerisk for misunderstandings based on divergent expectations and uses of terminology.

arken 3, 422 50 Hisings Backa,

son).

l rights reserved.

© 2009 Elsevier Ltd. All rights reserved.

1. Introduction

1.1. Historical background

Olof Kinberg, the first Professor of Forensic Psychiatry in Sweden,defined dangerousness as a “potential condition within a certainperson in a given environmental situation, which implies a particularrisk for a violation of other persons' legal rights or of the societalorganization itself” (Kinberg, 1955a). His standpoint was thatdangerousness was the result of the crime-eliciting forces (which hecalled “crime pressure”) being larger than the inner and outer crime-resistant forces (called “crime resistance”), and he introduced theconcepts of inner and outer static as well as dynamic risk factors(Kinberg, 1955b). His belief that “free will” was an illusion and thathuman behaviour was caused according to natural laws stronglyinfluenced his view of crime, criminal responsibility, and, not least,predictability. Kinberg advocated treatment rather than punishmentfor all perpetrators, and considered risk assessments to be reliable andreasonable grounds for treatment and rehabilitation efforts (Kinberg,1955b).

Kinberg's ideas had a great impact on Swedish legislation, andwhen the Penal Code of 1864 was replaced by the new Criminal Codeof 1965, the concept of “not guilty by reason of insanity” was aban-doned. In principle, all individuals convicted of a criminal act werejudged guilty, but prison sanctions were prohibited if the offenderwasfound to suffer from a major mental disorder (or similarly definedconditions) at the time of the crime. In most such cases, he or shewould be sentenced to compulsory forensic psychiatric inpatient care.A special discharge board, made up of an experienced judge, anindependent psychiatrist, further professionals as needed, and lay-men, had to approve applications for the discharge of forensicinpatients. Temporary leaves from forensic units could be revoked ifthe psychiatric status or behaviour of a patient showed signs ofdeterioration as judged by the psychiatrist in charge of treatment.

Up until the 1970's, risk assessments in Sweden, as in most othercountries, were regarded as clinical evaluations of dangerousnessprovided by psychiatry as an important service to society. Thischanged, however, when several international studies reported thatthe vast majority of individuals who had been assessed as “danger-ous” by psychiatrists and therefore deprived of liberty, did not relapsein criminality as predictedwhen released on legal grounds (Cocozza &Steadman, 1976; Steadman & Cocozza, 1974; Thornberry & Jacoby,1979). These findings, which, important to remember, have neverbeen disproved in new prospective studies, led to a dramatic

401T. Nilsson et al. / International Journal of Law and Psychiatry 32 (2009) 400–407

reconsideration of the value of such assessments. Psychiatric evalua-tions of the propensity to commit crimes rapidly fell into disreputeand had minor importance throughout the 1970's and 1980's inSweden as well as in the rest of the western world. A common opinionat the time was that “there are no dangerous patients — only dan-gerous situations”. In line with this, unlimited prison terms fordangerous criminals were also abolished as a sanction aftermodification of the law (SFS (Swedish Code of Statutes) 1979:680).As a consequence, the two forensic psychiatric units with the highestsecurity level in Sweden were closed.

1.2. Current situation

During the 1980's, a strong societal demand for coercive measuresagainst “dangerous”mentally disordered persons re-emerged, leadingto a new surge of interest in such predictions in the western world. InSweden, this was mirrored in new legal regulations of compulsorytreatment in general as well as in forensic psychiatry. The currentSwedish legislation (SFS 1991:1137) calls for assessments ofdangerousness if the perpetrator has committed a serious crime, ismentally disturbed, and sentenced to forensic psychiatric treatment. Ifa risk of relapse in serious criminality is assumed, the sentence toforensic psychiatric treatment involves “special court supervision”,which means that all decisions regarding leaves of absence anddischarge must be approved by an administrative court. Further,mentally disturbed inpatients must not be discharged from thehospital if judged to “constitute a risk” to himself or others. The lawprovides no definition of the risk to be assessed or of the time such anassessment is expected to cover. There is no similar demand for riskassessments in perpetrators sentenced to prison.

1.2.1. Forensic psychiatric investigationsThe majority of persons subjected to a forensic psychiatric

investigation in Sweden have committed some type of violent crime(Holmberg et al., 1999; National Board of Forensic Medicine, 2002),presumably under the influence of a mental disorder. Criminal identi-fication is generally not a predominant feature among these subjects,and they do not score especially high on assessment instruments suchas the PCL-R (Soderstrom, Sjodin, Carlstedt, & Forsman, 2004).

The vast majority (95%) of perpetrators who have undergone aforensic psychiatric investigation are assigned some type of psychi-atric diagnosis, but only about half of all investigated cases prove tohave a mental disorder severe enough to exclude a sentence to prison(Holmberg & Kristiansson, 2006; National Board of Forensic Medicine,2002). More than half of those exempted from regular sanctions sufferfrom psychoses, particularly of the type seen in the schizophreniaspectrum. In a smaller proportion of patients, the main diagnosis issevere compulsiveness or severe personality disorder with a tendencyto develop psychotic reactions in response to stress andmental strain.Finally, a few cases involve serious neuropsychiatric and develop-mental conditions characterized by considerable neurocognitivefunctional impairments. The majority of forensic psychiatric patientssubjected to risk assessments are thus perpetrators of a violent crimein association with a psychotic episode (National Board of ForensicMedicine, 2002).

1.2.2. Forensic psychiatric treatmentThe law of 1991 (SFS 1991:1129) implied that the risk for future

criminality must be assessed before a temporary or definite leavefrom forensic psychiatric units. Such decisionsmust inmost cases (the75% sentenced to treatment with “special court supervision” based onthe initial risk assessment, National Board of Forensic Medicine, 2002)be made by an administrative court.

For risk assessments during ongoing care, the main instructionissued by the National Board of Health andWelfare merely states thatthe care-giver shall provide “directions and ascertain that written

routines are at hand” (SOSFS 2006:9, p.6) for the determination of thepatient's propensity to hurt himself or others. The law text regulatingthe forensic psychiatric treatment explicitly states, however, thatwhen a patient is considered for less restricted conditions or whendischarge has been sought, the court shall evaluate the risk of criminalrecidivism along with the effects of the care and treatment the patienthas received (SFS 1991:1129). These evaluations are generally basedon the risk assessment made in connection with the initial courtproceedings combined with further information supplied by the care-giver. This means that persons in forensic psychiatric compulsory caremay come to undergo several formal risk assessments, e.g. whenabout to begin, change, or prolong periods of temporary leave, and inconnection with possible discharge.

In practice, more attention is given to the severity of the indexcrime than to the probability of a new offence, as exemplified by thefact that risk assessments usually are required before leave fromhospital in cases of murder but not in less fatal crimes, such as assaultand battery, although these crimes are associated with a much higherrisk of re-offending.

1.2.3. Long term prison inmatesForensically examined offenders found not to have a severemental

disorder prohibiting a prison sanction are in most cases diagnosedwith personality disorders, substance abuse disorders, and neuropsy-chiatric conditions such as attention deficit/hyperactivity disorder(AD/HD). Studies of general Swedish prison populations have alsorevealed high psychiatric morbidity (Holmberg et al., 1999; Levander,Svalenius, & Jensen, 1997) and high frequencies of neurocognitiveproblems such as AD/HD and dyslexia (Jensen, Lindgren, Meurling,Ingvar, & Levander, 1999). The subgroup of offenders with life-timesanctions is characterized by even higher frequencies of mentaldisorders than other convicts (Holmberg et al., 1999). Many have, incritical situations, committed crimes against persons they are close to.Few of these offenders have a pronounced criminal identification, andmost may be described in terms of “one-timers” (National Board ofForensic Medicine, 2006). This means that the life-timers stand out asrather atypical, and that they to a relatively large extent differ from theoffender category for which the risk prediction instruments havemainly been validated. Consequently, knowledge about risk assess-ment in this category is uncertain, and prisoners with life sentencesmay be grouped neither with severely mentally disordered nor withcriminally identified perpetrators (Manchak, Skeem,&Douglas, 2008).Despite this uncertainty, risk assessment of life-timers was mademandatory in 2006 (SFS 2006:45) in all cases given a hearing forpossible transformation to a fixed-time sentence. These assessmentsare performed by the National Board of Forensic Medicine by cross-disciplinary teams. In addition, precedential sentences by the SupremeCourt have stated that the results of these risk assessments should takepriority over other factors the court has to consider in order to reach adecision (Supreme Court Cases O 3016-07, O 2993-07, O 2182-07).

1.2.4. Official instructions and practical limitationsThe formal protocol of the risk assessments is partly regulated by law

and described in terms of directives and advice in official guidelinesissuedby theNational Board ofHealth andWelfare (SOSFS (theNationalBoard of Health and Welfare Code of Statutes) 1996:14; SOSFS 2006:9;SOFS 2006:16). It is stated that since single criteria most often haveweak predictive power on the individual level, “a comprehensiveanalysis and evaluation of the risk of relapse as well as the prospect ofreducing the risk shall bemade on the basis of all available information”(our translation).More detailed directives beyond these general recom-mendations on how to carry out risk assessments within a forensicpsychiatric investigation are not provided, and no attempt is made todifferentiate “mental” risk factors that requires contributions frommental health professionals from common-sense assessments based onassumptions such as “nothing predicts behaviour as behaviour”.

402 T. Nilsson et al. / International Journal of Law and Psychiatry 32 (2009) 400–407

1.2.5. Time framesOne aspect that is not at all touched upon in the legal texts,

instructions, or general advice is the question of the period of timethat the risk prediction is supposed to cover. Nothing is to be foundpertaining to this problem, neither in the sections concerning theforensic psychiatric coercive care nor in the documents addressing thequestion of time-limitation of life-time sentences. Though it is oftenclaimed that risk predictions are perishable goods with limited “bestbefore dates” (Borum, 1996), we know as little about how long anyrisk may persist as we know about how dynamic risk factors changeover time. Still, such assessments, and interpretations of such assess-ments, are likely to follow the patient through the years as non-questioned truths, just as “report parrots report” (Price, 1997).

As for shorter time frames, some risk factors, such as the risk ofimmediate harm to self and others, are routinely assessed in all(forensic) psychiatric consultations. Suicide risk is assessed on thebasis of historical or actuarial risk factors (e.g. sex, age, substanceabuse, previous suicide attempts, and recent significant losses) and thecurrent psychiatric status (e.g. depressive thought content, melan-cholic symptoms, suicidal ideation, and intense feelings of worthless-ness) (Bongar, Maris, Berman, Litman, & Silverman, 1993; Cardell,Bratcher, & Quinnett, 2009; Nordentoft, 2007). Immediate risk forviolence towards others may be assessed by following the sameprinciples. First to be considered are actuarial factors, such as previouscrimes, age, sex, and the presence of substance abuse and its relation topreviously violence. Next registered are any psychiatric factors thatmay be uniquely related to a transient phase of violent acting out in away that is causative from a clinical point of view, e.g. paranoidideation when there is a known potential victim, feelings of an immi-nent catastrophe during the first episode of schizophrenia, mixedmanic and depressive symptoms in affective psychosis, and extremeobsessive interests (Levander, 2000). The latter factors sometimesrequire a shorter period of involuntary treatment, for which there isnot only a civil law on compulsory psychiatric treatment but alsoeffective psychiatric treatment forms.

2. Scientific progress and practical problems

2.1. From dangerousness to risk

As we have seen from the beginning of the 1970s, assessments of“dangerousness” were in disrepute but resurfaced after some time as“risk” assessments, in what was called a new generation of violenceprediction research. This development was in tune with the emergingdominance of the risk concept in our culture from the beginning of the1980s (Beck, 2002). Instead of categorical assessments of “danger-ousness”, the “risk” of violence was measured as the proportion ofindividuals who relapsed or committed a certain type of crime in a(hypothetical) group sharing similar rating scores on structured orsemi-structured rating scales, or “instruments”, such as the “HistoricalClinical Risk management-20-items” (HCR-20) (Webster, Douglas,Eaves, & Hart, 1997; Webster & Eaves, 1995), the PsychopathyChecklist-Revised (PCL-R) (Hare, 1980, 1991), and the Violence RiskAppraisal Guide (VRAG) (Quinsey, Harris, Rice, & Cormier, 1998).These instruments were translated and adapted to Swedish condi-tions during the 1990's (Belfrage, Fransson, Söderberg, & Vasko, 1998;Grann, Långström, Tengström, & Stålenheim, 1998; Grann, Belfrage, &Tengström, 1998; Grann, Langstrom, Tengstrom, & Kullgren, 1999).

Assessments of dangerousness were formerly dichotomies likeyes–no, in–out, but risk assessments introduced probabilistic models(Steadman, 2000). The assessments are based on various sourcesof information, and the predictive validity of these sources varies.Historical and actuarial data maintain a robust predictive validity inpopulations of personality-disordered offenders, whereas clinical andrisk management factors may be of greater importance in offender

populations in which major mental disorders are prevalent (Litwack,2001).

2.2. Violence and risk assessment in the clinical situation

Comparatively little is known about clinical psychiatric risk factorsfor violent crime as the instruments listed have mostly relied onactuarial factors. Research on suicide risk assessments is far moreadvanced, and some lessons may undoubtedly be learned from thisarea. For both suicide and violent acting out, a number of factors maybe used to identify persons at increased risk. Conventional actuarialrisk assessments and epidemiological research may underestimatethe importance of specific psychiatric symptoms in the chain of eventsleading up to a crime.

Levander (2000) described the unique and disease-specific aspectsthat should be considered in the clinical situation. These risk aspects arenot the same as thosewe find in the combined clinical–actuarial instru-ments, such as the HCR-20, but rather unique and specific individualaspects of themental disorder, which, Levander claims, the experiencedclinician evaluates almost “intuitively”. The assessment is made withinthe frameof the clinical contact and focusedon the current status,wheredangerousness is a fairly rare phenomenon that is usually possible toevaluate only for a couple of days. Dangerousness is thus a continuousrather than a dichotomous factor. It is, for example, a well-knownclinical observation that active symptoms of psychotic illness withimperative hallucinations may temporarily increase the risk ofaggressive behaviour (Levander, 2000;Mullen, 1997). In schizophrenia,the risk of violence may be increased during the first acute phase of theillness (possibly due to confusion) and in the later phases when severeviolence is more often directed toward close relatives, e.g. parents, thantowards strangers (Nordström, 2005).

2.3. Methodological problems

The psychiatric risk assessment methods developed during the lastdecade are not without technical problems. Mossman (1994) hasillustrated how the predictive value and the proportion of correct andincorrect assessments may vary between raters even when using afictitious instrument with high predictive ability and a specifiedprevalence of relapses that does not differ between groups.

The Receiver Operating Characteristics (ROC) has become thestandard measure of predictive ability. This method is well suited forsolving two kinds of problems in research on risk prediction. First, thecut-off score that is most effective in distinguishing signal from noise,i.e. relapse from non-relapse, can be calculated from the curve's pointof inflection (Rice & Harris, 1995), and, second, the area under thecurve (AUC) can be used as a measure of the instrument's overallcapacity to distinguish between signal and noise at all possible cut-offscores (Mossman, 1994, Green & Swets, 1966). This value may also beinterpreted as the probability of an actual signal having a higher scorethan the noise in the continuous variable (Rice & Harris, 1995).

It may be edifying to dwell upon the kinds of information theseanalyses can and cannot provide. ROC analyses were invented as amethod with the specific aim of finding the most effective cut-offscore between signal and noise in complex information (Pepe, 2000;Zweig & Campbell, 1993). The AUC is thus very useful in comparingthe theoretical accuracy of different test methods at all possible cut-off levels, but provides no information about the accuracy in a singleprediction, which would require a predetermined cut-off and knowl-edge of the prevalence of the studied phenomenon in the actual groupof subjects. The use of ROC to evaluate a research approach wherecategorical statements about dangerous/non-dangerous are avoidedin favour of estimated degrees of risk is also fundamentally contra-dictory, since ROC was developed as a method for transforming con-tinuous data to categorical data (Fawcett, 2006).

403T. Nilsson et al. / International Journal of Law and Psychiatry 32 (2009) 400–407

A further methodological aspect that needs to be addressed beforeintroducing new methods is the reliability, especially when usingmany different sources of information clinically, across raters and overtime. A greater proportion of inaccurate ratings can be expected if themethods are used by others than research teams trained in co-rating.There are no published studies of co-rating in ecologically validsettings. This would hardly be regarded as an acceptable basis fordecisions affecting individuals in any other field of medicine, and it isvery far from the degree of reliability required to convict somebody ofa crime.

2.4. Risk management

A “third generation” of risk research has now been proposed as “riskmanagement”, including a structured clinical assessment focusing riskfactors in order to manage individual cases, instead of performingpredictions. The obvious argument against this is that since risk factorsare not defined as causal factors, it cannot be taken for granted thatmanagement of such factors necessarily influences the outcome. Astructured professional judgment should also consider availability andacceptance of support after discharge (National Board of ForensicMedicine, 2000), which may be difficult to assess in advance.

2.5. Recent research: limitations, problematic areas,and possible solutions

Recent years have brought research that has shed further light onthe difficulties and uncertainties surrounding risk assessments ofindividual patients. One such problem is the transition from thenomothetic to the idiographic level. Practically all research has beennomothetical and thus dealt with number of relapsing individuals in agroup without attention to the single individual, i.e. the ability toidentify the individualswho actually relapse. In an analysis of actuarialrisk assessment instruments (ARAIs), Hart, Michie, and Cooke (2007)have shown that on the individual level, the confidence interval for thesingle individual becomes too large to meaningfully express anythingabout his risk of relapse. ARAIs can therefore not be used to estimate anindividual's risk for future violence with any reasonable degree ofcertainty and should therefore be used with great caution or not at all.Since assessment instruments like the HCR-20 and PCL-R have notbeen shown to be superior at identifying the individual who willrelapse, they should be seen as a complement and not a substitute to acareful and clinically informed appraisal (Dahle, 2006).

The concept of psychopathy as reflected in the PCL-R has long beenseenasa corepersonality indicator of riskacrossmethods anddiagnosticcategories. Recent research has somewhat changed this picture, since itseems to be the behavioural factor, i.e. previous criminal behaviour,rather than the personality facets of the psychopathic construct thatconstitutes the predictive power of PCL-R. Walters, Raymond, Grann,and Dahle (2008) have in a collaborative study analyzed six samplesfrom different countries with respect to the four factors of PCL-R, wherethey found virtually no support for any incremental validity of factors 1,2, and 3 (Interpersonal, Affective, and Lifestyle, respectively) above andbeyond factor 4 (Antisocial behaviour).

As the supervising authority for specialized psychiatric care inSweden, the National Board of Health andWelfare is also the authoritywith the ultimate responsibility for the quality aspects, including thescientific evidence for the use of risk assessments within this area. Inthe wake of several incidents that attracted public attention in theearly 2000s, the National Board of Health and Welfare decided toattend to the question of risk assessment in forensics as well as ingeneral psychiatry.

In 2002, the Board published a follow-up study, which includedall patients in forensic psychiatric care on September 27, 1995 andfollowed them for about 3 1/2years through the National Police Boardregisters (National Board of Health and Welfare, 2002). The total

number of patients came to 665 (8% of whom were women), and 309patients had been discharged by the end of the study period. Anunexpected finding was that 28% of the patients were convicted ofnew crimes during an ongoing treatment, and especially remarkablewas that 30 (5%) of these offenders had committed serious crimes ofviolence (sexual crimes in 19 cases, manslaughter in 8, homicide in 7,and aggravated assault in 9). Sixty-five of the 309 patients who weredischarged during the follow-up period (21%) relapsed in some kindof criminality, including 38 persons (12%) who relapsed into violentcrimes. The only factor found to be significantly correlated to relapsein violent criminality was an ongoing substance abuse at discharge.The high relapse rates, both during ongoing care and after discharge,indicate a poor ability to identify relapse propensity in combinationwith a failure to deal with significant risk factors such as ongoingsubstance abuse (National Board of Health and Welfare, 2002).

2.6. “Insane acts” and their consequences for risk assessment

A number of severe and very high-profile crimes that occurredduring 2003 came to turn the attention of a shocked general public tooffenders with mental disorders. The murder of the Foreign SecretaryAnna Lindh had an enormousmedia impact and was soon followed bya couple of other incidents referred to as “insane deeds” in the press. Acommon factor among these perpetrators was that they had soughtpsychiatric care shortly before the crimes but not been admitted toinpatient treatment or prematurely discharged. It was against thisbackground that the National Board of Health and Welfare was com-missioned by the government to 1) make a survey of the use of riskassessments in the specialized psychiatric care section, and 2) compilea review on the scientific evidence for risk assessments in cooperationwith the Swedish Council on Technology Assessment in HealthCare. The first task resulted in the report “Risk assessment withinspecialized psychiatric care” (National Board of Health and Welfare,2004).

Summarized in thefinal chapter of this report are the current praxisas well as present knowledge in the field. Given the Board's status as asupervisory agency, this text must be regarded as an instruction. Forthis reason, it seems extremely problematic that the directives arecontradictory and vague, not least since risk assessments in manycases are regulated by law (SFS 1991:1129). By way of introduction,risk assessments are described as “fresh goods”, where the currentlevel of knowledge permits no “psychiatry-based” determinations ofrisk of violent acts beyond the immediate perspective in the presenceof amental disorder requiring treatment (National Board ofHealth andWelfare, 2004, p. 31). It is hard to interpret this declaration in any wayother than that of the question of risk of violent acts can only beelucidated in the very brief and directly disorder-related perspective.In the next paragraph, the Board appears to advocate a moderate andcautious strategy for risk assessments, as the importance of notmisjudging the risk is stressed, i.e. that preventivemeasures should bebased on a threshold value which to the greatest possible extentattempts to avoid false negatives while trying to keep down thenumber of false positives, so that measures involving deprivation ofliberty will not be used for preventive purposes (National Board ofHealth and Welfare, 2004, p. 31). It is nevertheless difficult to upholdsuch a strategy without placing a rather large proportion of subjectswho would not have behaved violently among those judged as“dangerous” (Grann & Nilstun, 2000).

This impression is also supported by the reference to article 5 ofthe European Convention and to the Madrid Declaration, both ofwhich are documents that clearly state that psychiatry must not “beused to detain individuals on the grounds of perceived dangerous-ness” (National Board of Health and Welfare, 2004, p. 31). When theauthors go on to discuss risk assessments as “one link in a whole chainof measures” that are all aimed at reducing risk behaviour, i.e. the riskassessment procedure is described as a process with the aim of

404 T. Nilsson et al. / International Journal of Law and Psychiatry 32 (2009) 400–407

“preventing the occurrence of identified risks”, it becomes difficult tounderstand the intention of the instruction (National Board of Healthand Welfare, 2004, pp. 32–33). And it becomes even more difficultwhen it is declared that “the risk that the patients commit violentcrimes against others must be as methodically evaluated as we todaydetermine the risk of suicide” (National Board of Health and Welfare,2004, p. 33). With this shift in the argumentation, it seems that theBoard has put one foot firmly in each camp, first stating that thescientific support for risk assessments is limited and calls for greatcaution, and then goes on to argue that risk assessments shouldalways be carried out regarding the potential risk of psychiatricpatients to commit violent crimes against others.

2.7. Evidence-based knowledge and risk assessment

The second task given to the National Board of Health andWelfarein collaboration with the Swedish Council on Technology Assessmentin Health Care (Statens Beredning för Medicinsk Utvärdering (SBU))was to examine the scientific support for the methods currently em-ployed for assessing the risk that a psychiatric or forensic psychiatricpatient one day will commit an act of violence. Their conclusions werepresented in the report entitled “Risk assessment within psychiatry.Can violence in society be predicted?” (our translation) (SBU, report175, 2005). The initial step of the investigation was a PubMed searchfor literature (published after 1993) with wide criteria to find asmanypotentially relevant studies as possible, followedby a PsychInfo search,which altogether generated 4052 articles. The aim was to establishfigures for the accuracy of risk assessments of psychiatric in- and out-patients from both general and forensic psychiatry. Studies on personsadmitted to hospital for observation or investigation regardingcriminal behaviour were included as were the studies on pre-trialforensic investigations if it was clear that mental disease, dementia,drug abuse, mental retardation, or personality disorders were amongthe main reasons for the investigation. Studies including personsadmitted to psychiatricwards in the prisonswere included, but studiesof subjects without mental problems who were incarcerated due tocriminality and assessed for risk of violent recidivismwere not. Studiesof mixed groups (person in prison and special hospital) were includedif the prison inmates had clearly defined mental problems, such aspersonality disorders, drug abuse, and mental retardation. Afterexcluding studies examining the risk for violence inside psychiatricwards or other psychiatric settings, the review was finally limited tostudies focused on risk for violence in society. Meta-analyses andliterature reviews were compiled and reported separately. Carefulevaluation according to strict criteria by two researchers indepen-dently of each other resulted in 37 relevant articles, 27 of whichconcerned risk assessment of forensic psychiatric patients, i.e. personsunder care because they had committed serious crimes. The remainingten studies involvedgeneral psychiatric patients. Eachof the37 studieswas assessed for scientific value by an instrument containing ninecriteria, each weighted on a three-graded scale (high, medium, or lowscientific evidential value). The final evidential value was based on acombined estimation of all the criteria in which design and size ofsample (n) were more important than the others, giving the highestgrades to studies with n>200 and the lowest to studies with n<50.

The results confirmed that psychiatric risk assessments weresuperior to chance in identifying the patientswhowould subsequentlyengage in violence. Therewas support for the value of risk assessmentsin both forensic and general psychiatry. Although forensic psychiatricpopulations proved to be studied more frequently, the scientific evi-dence was stronger in the general psychiatric studies and thus rathersupported the practice of risk assessments among general psychiatricpatients in comparison to forensic psychiatric patients. Risk assess-ments based on currently available and evaluated methods, indepen-dent of how carefully accomplished, were also always connected witha certain measure of uncertainty reaching a margin of error with at

least 25–30%, i.e. every third or fourth case aremisjudged.Women andethnic minorities were not represented satisfactorily, and most studypopulations were small and not representative for a nation or largerregions, particularly not in Sweden. Neither did any study support theclaims for validity of forensic psychiatric risk assessments with a “highvalue of evidence”. All in all, the results of the literature survey did notallow further general conclusions than those outlined above.

3. Ethical implications

3.1. Medical ethics and risk assessments

Risk analysis applied to any field has a wide range of ethical im-plications (Lewens, 2007). Central to forensic psychiatry is the balancebetween conflicting individual interests, professional values, thepublic interest of safety, and important social values, such as justice.Forensic psychiatric risk assessments take place in a medical, andusually institutional, context, where the responsibility of using well-confirmed methods in a way that does not generate (by misunder-standing ormisuse) too serious adverse primary, secondary, or tertiaryeffects, lies on medical scientists and practitioners (Beauchamp &Childress, 2001). The core of the basic ideas expressed by standardmedical research ethics declarations since the Nuremberg Code (cf.Council for International Organizations of Medical Sciences (CIOMS),2002) is that there is a limit to what sort of socio-political schemesmedical professionals can participate in without acting unethically.

The scientific concept of risk is an aggregate of an undesirableoutcome and the probability that this outcome will actually occur.Since forensic psychiatric findings do not allow distinct discriminationbetween different sorts of crimes regarding the probability of relapse,it is doubtful if forensic psychiatric science can actually contribute to arisk assessment from a general point of view.Moreover, since society'sperspective on the severity of different crimes and the evaluation ofdifferent classes of offenders may be questioned on ethical grounds, itmay not be compatiblewithmedical ethics to take an active part in thelegal system. Thus, the old issue of the responsibility of psychiatricexpertise in a wider social context seems to be as alive today as it wasin the 1970's, when the initial opposition to assessments of danger-ousness was loud and clear. In view of professional ethics as well asgeneral legal and social values, it is particularly crucial to consider ifforensic psychiatric risk assessments contribute to prejudicial andunfair discrimination of people with mental health problems in thelegal system.

None of this would imply, of course, that forensic psychiatric riskassessment models may not be of clinical or scientific value whenused, for instance, in the context of evaluation and prioritization oftreatment strategies or research hypotheses in relation to differentgroups of patients or subjects. In that case, to employ themethods andmodels of risk assessment will more or less equal the making ofcareful, evidence-based clinical decisions. However, from an ethicalperspective continuous evaluation of outcomes in clinical terms inorder to improve the evidence base implied by this will not by itselfhave any implication for the probability of recidivism, but rather forthat of the occurrence of clinical symptoms given different treatmentstrategies and different patient groups.

In other words, the ethical challenges have to do with the way theinformation provided by forensic psychiatric risk assessments isunderstood, not by forensic psychiatrists but by the legal system andthe public. The more weight the legal system puts on a particulargeneral presumption regarding persons with mental health problemscompared to others when it comes to the likelihood of criminalbehaviour (particularly, the likelihood of offenders with such prob-lems to relapse into crime as compared to mentally less unhealthyoffenders of comparable crimes), the more can such a presumption beexpected to be reflected in the popular opinion. If the collectedknowledge within forensic psychiatry cannot lend support to such a

2 In Munthe, Radovic, and Anckarsäter (in press), it is argued that a similar point canbe made also regarding forensic psychiatric researchers and standard research ethicalnorms.

3 A radical alternative, suggested by Appelbaum (1997), is to hold out the forensic

405T. Nilsson et al. / International Journal of Law and Psychiatry 32 (2009) 400–407

presumption, itmay be seen as questionable to participate in a practiceof forensic psychiatric risk assessment where the results will beinterpreted on the basis of such a priori assessments. From themedicalethics perspective, the emphasis should in that case rather be oncounteracting what, in the light of available evidence, looks likeunsubstantiated prejudice against persons with mental healthproblems. It is not certain that just insisting on scientific validity andunderscoring uncertainties in court presentations etcetera suffices toaccomplish this. Not even the very defensive strategy, suggested byAppelbaum, to approach the task as merely giving a “factual back-ground and interpretative context” for the assessments and decisionsof other people1 may be enough.

Below, it will be demonstrated that most legal systems in factoperate on the basis of a (sometimes implicit) presumption of the sortjust described, and that this constitutes a severe lack in terms ofjustice. The gist of the problem is this: forensic psychiatric riskassessments made to inform the court, in such a context, serve as anaffirmation of a presumption that people who commit crimes underthe influence of mental disorders are generally more dangerous thanother offenders. The presence of such a presumption is evidentwhenever it is a built-in feature of the legal system to require more ofmentally disordered offenders for these to escape incarceration thanfor other offenders of comparable crimes. For instance, it is a wide-spread practice in many jurisdictions (among these, Sweden) torequire a favourable forensic psychiatric risk assessment as a condi-tion for discharge from a forensic psychiatric hospital, besides suchconditions that are applied also to prison inmates regarding questionsabout parole, release, etcetera. At the same time, as has been seenabove in the review of the scientific development, besides such pre-dictive factors thatmay be applied equallywell to all sorts of offendersregardless of their mental health status (such as historical patterns ofcriminal behaviour, drug abuse, gender, or socio-economic status),auxiliary information about mental health has a very uncertain–atleast unsubstantiated–added predictive value regarding future crim-inal behaviour. Thus, no matter how carefully they are expressed, theresults of forensic psychiatric risk assessmentswill as a rule be fed intoa system where mentally disordered offenders are systematicallytreated less favourably than other offenders on grounds that lacksupport from the very same scientific point of view that informs theserisk assessments.

The risk assessor is inmany cases also the care-giver (physician) ofthe person assessed. This is important for securing quality in judg-ments based on clinical experience, for achieving correct classificationof the patient's condition when making a bona fide risk analysis. As aclinical care-giver, the professional is supposed to act on the bestinterest of his patient. For a forensic psychiatric risk assessor, however,it is the public interest that is in focus, and thismaynot always coincidewith the patient's best interest. The opposing practical “pulls” of theseroles in such casesmay influence each other, and the decisionwill be inconflict with one or the other of the aims of these two professionalroles and thus constitute an ethical dilemma (Robertson & Walter,2008). What is more, this factor introduces a further source ofuncertainty into the legal system. The risk assessor may, for instance,be more influenced by concern for his patient than by the need toprotect potential victims — or vice versa.

There are thus two main basic ethical problems — one concerningprofessional ethics and one concerning social justice. If there is no, orvery poor, scientific basis for forensic psychiatric predictions of futurecriminality, and if it is possible or even probable that the informationdelivered will be misused, misunderstood or in other ways sys-tematically favour an outcome that is not in the patient's interest,there are standard views in medical ethics suggesting that the pro-

1 Appelbaum, (2008), note 1, p. 195.

fessional clinician2 should decline to answer society's call for riskassessments.3

A more nuanced solution might be to approach the issue on a caseby case basis — but to decline when neither science nor clinicalexperiences indicate that forensic psychiatric factors wouldn't add tothe overall prediction. In that case, the professional can deliversomething approaching what society asks for whenever it is notirresponsible (if done in the right way) from a professional, ethicalpoint of view. This solution presumes that delivering risk assessmentsis not–as it currently is in Sweden–a systematic requirement bysociety laid on professional forensic psychiatric expertise.

Looking at the institution of health care as a whole adds a furthersource of criticism to the one based on the particular relation betweenan individual patient and an individual risk assessor. Tännsjö (1999,chap. 6) has claimed that the institutional goals of health care (asexpressed in medical ethical principles and values) are too farremoved from the goals of the legal and penal system to involvemedical professionals in legal decisions regarding criminal justice.This turns the focus to the ethical problem of social justice. It may beargued against Tännsjö that the overall importance of the valuesprotected by the legal system in this respect are of such weight thatthey should sometimes take precedence over professional ethicalvalues even at an institutional level.

In areas such as public health, medical professionals partake inpolicies where the best interest of patients is sometimes sacrificed forthe sake of protecting overarching social values, such as stability andwelfare (Dawson, 2008;Munthe 2008). Thus, theremay be some roomin medical ethics for forensic psychiatric experts to participate in theexecution of criminal justice for similar reasons, especially if there areways to solve the professional ethical dilemma involved in forensicpsychiatric risk assessment.4 This mode of reasoning rests on thepremise that the criminal justice system is just — that it really isdesigned to protect important social values. However, the observa-tionsmade above regarding the use that ismade of forensic psychiatricrisk assessment in many jurisdictions cast some doubt on thatassumption. This issue, therefore, deserves some further examination.

3.2. The legal system and the mentally disordered offenders

As pointed out above, the common presumption that mentallydisordered offenders aremore dangerous than other people in generalor, in particular, other perpetrators of comparable crimes, rests on ashaky scientific ground. When this presumption is reflected in asystematically unequal treatment with regard to the incarceration ofmentally disordered and other offenders, it takes on the shape of anunfounded prejudice giving rise to unjust discrimination.

From the point of view of society, it may, of course, seem apt toapply some sort of precautionary strategy when it comes to the legalevaluation done on the basis of a forensic psychiatric risk assessment.That is, even if there is no, or very weak, scientific evidence for thebelief that this particular mentally ill individual is more likely thansomeone else to commit a crime, and even if we are unable to establishany causal connection between particular risk factors and specificoutcomes, it may make sense to apply the rule of “better safe thansorry”. For example, if the person has committed several crimes of asimilar type before and on top of that ismentally unstable in away that

psychiatric profession as serving solely the criminal justice system, thus not beingsubject to standard requirements of medical ethics. This view has not gained widesupport in the forensic psychiatric community, however.

4 This brief sketch of an analysis is developed in greater depth in Munthe, Radovic,and Anckarsäter (in press).

406 T. Nilsson et al. / International Journal of Law and Psychiatry 32 (2009) 400–407

may be suspected to possibly affect the likelihood of relapse, let us playit safe and keep him isolated from society! Such reasoningmay appearparticularly attractive when it comes to very severe criminality.

However, precaution works both ways (Sandin, 1999). Just as wemay find the “better safe than sorry” approach attractive applied tothe possible occurrence of crime, we would surely like to apply thesame principle to the possible occurrence of systematic discriminationof certain categories in the legal system. All civilized nations do atrade-off between the public interest of fighting crime and the publicinterest of upholding decent standards of judicial security, where thelatter is given a very high significance. So, the precautionary reasoningmay as well be applied to the other side of that equation, thus arguingthat in order not to jeopardize judicial security, in want of solidevidence to the contrary, mentally ill offenders should not be treatedas a priori more likely to relapse in a serious crime than otheroffenders. Thus, applying forensic risk assessment in decision makingabout release or discharge, using factors such as criminal history, maybe quite acceptable as long as the precautionary reasoning appliedbased on such an assessment treats all offenders with a burdeningcriminal history of serious crime alike. This, however, is not thecommon practice in most jurisdictions — where psychiatric risk isroutinely added on top of the well-confirmed forensic risk factors.

From the point of view of ethics and forensic psychiatric science,the provocative conclusion seems to be that there are strong reasonsto question any legal system built on the presumption that mentallydisordered offenders are especially dangerous in a way that motivatesfurther incarceration on purely mental health grounds, compared toother offenders of comparable crimes and with a comparable generalrisk of recidivism. To be clear, the risk of relapse can thus be assessedfor mentally disordered offenders as it can for any offender, but thereare strong reasons to avoid building into the legal system thescientifically unproven (if not disproved) prejudice that mentally illpeople who have committed crimes are generally and substantiallymore likely to relapse into criminal behaviour than comparableoffenders just because of their mental health status.

In sum, careless use of uncertain psychiatric data in this fieldthreatens to jeopardize judicial security, mislead public attention, andreinforce prejudice against people with mental health problemsamong the general public. While the use of forensic psychiatric riskassessment may be well motivated as a tool in the clinical or researchsetting, it is for the reasons just explained much more uncertain if thepractice of contributing such assessments to legal processes of the sortdescribed above can be defended from the standpoint of medicalethics. At the same time, in Sweden and several other countries,forensic psychiatric expertise is expected or even obliged to performexactly such tasks. As a consequence, medical ethical responsibilitieswould seem to conflict not only with widespread practices as regardforensic psychiatric risk assessment, but with professional responsi-bilities tied to the role as a public official as well.

4. Concluding remarks

On the whole, the progress of risk assessments in Sweden duringthe last decade has followed the international development and isnowadays carried out as a structured clinical assessmentwith Swedishversions of instruments such as HCR-20 and PCL-R. By these methods,subject groups with a higher probability than others for violent actingout, especially re-acting out, can in fact be identified (SBU, 2005), but itis much more difficult to assess individuals and to arrive at a reason-able conclusion about the individual risk of relapsing into criminality(Hart et al., 2007). These methods also require special training inpsychiatry or psychology, and thereby implicate that the assessmentsmade are based on somethingwhich is psyche-related in quality. Clearaccounts of previous behaviour, for us the most important predictorfor new crimes, are simply found in the criminal records, whetheror not the offenders are mentally disordered and do not require the

assessor to be trained either as a psychiatrist or a psychologist. Malegender, youth, substance abuse, and low age at onset of antisocialbehaviour always turn up as risk factors in research on the probabilityof relapse in violence (Eriksson, 2008; Hastings & Hamberger, 1997).The relatively wide variation in the current forms and use of riskassessments and risk prevention described in the inspection of thespecialized psychiatric care (National Board of Health and Welfare,2004), from systematic andwell structuredmeasures tomore sporadicand unstructured efforts, can most likely be explained by the fact thatthe overall guidelines have been unclear and full of contradictionsalong with the fact that no assessment method is more accurate thanany other.

Rather than issuing contradictory statements, the National Boardof Health and Welfare as the authority in charge should have made asystematic cost–benefit analysis regarding the use of risk assessmentsand the measures based on such assessments that could reasonablyplay a role within psychiatric care. The characteristics of the riskassessment instruments are relatively well-known today, both whenit comes to accuracy and margins of error, such as number needed todetain in order to prevent a relapse in crime, our access to scientificinformation is more limited (Buchanan & Leese, 2001). As in othermedical areas, this field needs development towards a praxis based onanalyses of economic as well as human costs, and of ethical aspects aswell as positive and negative effects. By initiating such analyses, theauthorities involved would contribute to the establishment of anobjective and scientific basis for a broad discussion on the politics ofcare, which could guide politicians in taking a stand on in whichcontexts care givers should be engaged in risk assessments and whichpreventive measures the assessments should lead to.

A basic demand on just legislation is that all offenders are to betreated equally and fairly,which ishardly the case judging fromthewaysociety has singled out the category of mentally disordered subjects asespecially perilous. They are supposed to be extensively scrutinisedand, when there is a risk for relapse into criminality, they are handedover to an unlimited form of detention with considerably reducedindividual rights. This focus on the mentally disordered offenders isinitiated by politicians and policy makers, but also by professionalsclaiming expertise and authority in the field, and is written into thelaws that regulate the forensic psychiatric practice. There is thus animbalance between regular offenders and those who suffer from asevere mental disorder that is difficult to justify. Despite thisremarkable focus on the mentally disordered offenders which is likelyto entail repressivemeasures andpotential integrity violations, Swedenhas had very little political debate questioning this development.Instead, the supervising authorities have increasingly emphasized theimportance of performing risk assessments at the same time as thescientific evidence for these assessments has been disputed, lendingvery little support to the psychiatrists and psychologists supposed tocarry out the assessments. The idea of special rules for certaincategories of people is not rational and might jeopardize establishedrelations with patients as well as the general confidence in our work.

What we now need is research to obtain specific knowledge aboutall types of risk factors and all types of offenders in specified timeframes, and the explanatory value of all types of factors that maypredict crimes and relapse in crimes, be they psychiatric, behavioural,socio-economic or cultural. This information then has to be weighedagainst the usefulness of predictions in order to reduce the risk forcrimes, and evidence-based strategies to do so. Only then will thelegislators have the information they need for the political process ofcreating a rational penal law.

References

Appelbaum, P. S. (1997). A theory of ethics for forensic psychiatry. Journal of theAmerican Academy of Psychiatry and the Law, 25, 233−247.

Appelbaum, P. S. (2008). Ethics and forensic psychiatry: Translating principles intopractice. Journal of the American Academy of Psychiatry and the Law, 36, 195−200.

407T. Nilsson et al. / International Journal of Law and Psychiatry 32 (2009) 400–407

Beauchamp, T. L., & Childress, J. F. (2001). Principles of biomedical ethics. Oxford: OxfordUniversity Press.

Beck, U. (2002). Risk society: Towards a newmodernity, theory, culture & society. London:Sage Publications.

Belfrage, H., Fransson, G., Söderberg, E., & Vasko, T. (1998). Rättspsykiatriprojekt i Växjöoch Sundsvall: Nytt skattningsinstrument ger samstämmiga riskbedömningar.Läkartidningen, 95, 2469−2473.

Bongar, B., Maris, R. W., Berman, A. L., Litman, R. E., & Silverman, M. M. (1993). Inpatientstandards of care and the suicidal patient. Part I: General clinical formulations andlegal considerations. Suicide and Life Threatening Behavior, 23, 245−256.

Borum, R. (1996). Improving the clinical practice of violence risk assessment. AmericanPsychologist, 51, 945−956.

Buchanan, A., & Leese, M. (2001). Detention of people with dangerous severepersonality disorders: A systematic review. Lancet, 358, 1955−1959.

Cardell, R., Bratcher, K. S., & Quinnett, P. (2009). Revisiting “suicide proofing” an inpatientunit through environmental safeguards: A review. Perspectives in Psychiatric Care, 45,36−44.

Cocozza, J., & Steadman, H. (1976). The failure of psychiatric predictions of dangerousness:Clear and convincing evidence. Rutgers Law Review, 29, 1084−1101.

Council for International Organizations of Medical Sciences (CIOMS). (2002). Interna-tional ethical guidelines for biomedical research involving human subjects. Geneva:CIOMS.

Dahle, K. P. (2006). Strengths and limitations of actuarial prediction of criminalreoffence in a German prison sample: A comparative study of LSI-R, HCR-20 andPCL-R. International Journal of Law and Psychiatry, 29, 431−432.

Dawson, A. (Ed.). (2008). The philosophy of public health. Aldershot: Ashgate.Eriksson, Å. (2008). Schizophrenia and criminal offending: Risk factors and the role of

treatment. Doctoral Thesis. Stockholm: Department of Clinical Neurosciences,Karolinska Institutet.

Fawcett, T. (2006). An introduction to ROC analysis. Pattern Recognition Letters, 27,861−874.

Grann, M., Belfrage, H., & Tengström, A. (1998). Actuarial assessment of risk of violence:predictive validity of the VRAG and the historical part of the HCR-20. In: M. Grann(thesis), Personality disorder and violent criminality. Stockholm: KarolinskaInstitute.

Grann, M., Långström, N., Tengström, A., & Stålenheim, E. G. (1998). Reliability of file-based retrospective ratings of psychopathy with the PCL-R. Journal of PersonalityAssessment, 70, 416−426.

Grann, M., Långström, N., Tengström, A., & Kullgren, G. (1999). Psychopathy (PCL-R)predicts violent recidivism among criminal offenders with personality disorders inSweden. Law and Human Behavior, 23, 205−217.

Grann, M., & Nilstun, T. (2000). Rättspsykiatriska riskbedömningar är etiskt försvarbara.(Forensic psychiatric risk assessment is ethically defensible). Läkartidningen, 97,581−583.

Green, D. M., & Swets, J. M. (1966). Signal detection theory and psychophysics. New York:John Wiley and Sons Inc.

Hare, R. D. (1980). A research scale for the assessment of psychopathy in criminalpopulations. Journal of Personality and Individual Differences, 1, 111−119.

Hare, R. D. (1991). Manual for the Hare psychopathy checklist-revised. Toronto: Multi-Health Systems.

Hart, S. D., Michie, C., & Cooke, D. J. (2007). Precision of actuarial risk assessmentinstruments. British Journal of Psychiatry, 190, 60−65.

Hastings, J. E., & Hamberger, L. K. (1997). Sociodemographic predictors of violence.Psychiatric Clinics of North America, 20, 323−335.

Holmberg, G., Forsman, A., Grann, M., Ingerloo, L. -E., Skagerberg, S., & Somander, L.(1999). Psykiatrisk vård för fängelsedömda. (Psychiatric care for those sentencedto prison). Nordisk Tidskrift for Kriminalvidenskap, 86, 206−219.

Holmberg, G., & Kristiansson, M. (2006). Contacts with public services, with specialreference to mental health care, preceding a serious crime: A retrospective study of268 subjects of forensic psychiatric investigations. International Journal of Law andPsychiatry, 29, 281−288.

Jensen, J., Lindgren, M., Meurling, A. W., Ingvar, D. H., & Levander, S. (1999). Dyslexiaamong Swedish prison inmates in relation to neuropsychology and personality.Journal of International Neuropsychological Society, 5, 452−461.

Kinberg, O. (1955). Kriminologiska grundproblem. Stockholm: Wahlström & Widstrand.Kinberg, O. (Ed.). (1955). Samhället och brottslingen (pp. 23−60). Stockholm:

Norstedts.Levander, S. (2000). Klinisk farlighetsbedömning. (Clinical assessment of dangerous-

ness). National Board of Forensic Medicine. Riskbedömning vid rättspsykiatriskundersökning — Riktlinjer och reflexioner. (Risk assessment during forensic psychiatricinvestigations — General outlines and reflections). Stockholm: RMV-report 2000:1.

Levander, S., Svalenius, H., & Jensen, J. (1997). Alarming figures on inmates at Swedishprisons. Mental disorders are common among the inmates. Läkartidningen, 94,46−50.

Lewens, T. (Ed.). (2007). Risk: Philosophical perspectives. London: Routledge.Litwack, T. R. (2001). Actuarial versus clinical assessments of dangerousness. Psychol-

ogy, Public Policy, and Law, 7, 409−443.Manchak, S. P., Skeem, J. L., & Douglas, K. S. (2008). Utility of the Revised Level of Service

Inventory (LSI-R) in predicting recidivism after long-term incarceration. Law andHuman Behavior, 32, 477−488.

Mossman, D. (1994). Assessing predictions of violence: Being accurate about accuracy.Journal of Consulting and Clinical Psychology, 62, 783−792.

Mullen, P. E. (1997). A reassessment of the link between mental disorder and violentbehaviour, and its implications for clinical practice. Australian and New ZealandJournal of Psychiatry, 31, 3−11.

Munthe, C. (2008). The goals of public health: an integrated multidimensional model.Public Health Ethics, 1, 39−52.

Munthe, C., Radovic, S., & Anckarsäter, H. (in press). Ethical issues in forensic psychiatricresearch on mentally disordered offenders. Bioethics.

National Board of Forensic Medicine (2000). Riskbedömning vid rättspsykiatriskundersökning — Riktlinjer och reflexioner. (Risk assessment during forensicpsychiatric investigations — General outlines and reflections). Stockholm: RMV-report 2000:1.

National Board of Forensic Medicine. (2002). Annual statistics. http://www.rmv.se/National Board of Forensic Medicine (2006). National inventory of life-timers. Working

paper.National Board of Health and Welfare (2002). Rättspsykiatrisk vård. Utvärdering–

omvärdering. (Forensic psychiatric care. Evaluation–revaluation) Stockholm:National Board of Health and Welfare.

National Board of Health and Welfare (2004). Riskbedömningar inom den specialiser-ade psykiatriska vården. En kartläggning. (Risk assessments within the specializedpsychiatric care. An overview) Stockholm: National Board of Health and Welfare.

Nordentoft, M. (2007). Prevention of suicide and attempted suicide in Denmark.Epidemiological studies of suicide and intervention studies in selected risk groups.Danish Medical Bulletin, 54, 306−369.

Nordström, A. (2005). Violent offenders with schizophrenia. Quantitative and quali-tative studies focusing on the family of origin. Doctoral Thesis. Umeå: Departmentof Clinical Sciences, University of Umeå.

Pepe, M. S. (2000). Receiver operating characteristic methodology. Journal of theAmerican Statistical Association, 95, 308−311.

Price, R. (1997). On the risks of risk prediction. Journal of Forensic Psychiatry, 8, 1−4.Quinsey, V. L., Harris, G. T., Rice, M. E., & Cormier, C. (1998). Violent offenders: Appraising

and managing risk. Washington D.C: American Psychological Association.Rice, M. E., & Harris, G. T. (1995). Violent recidivism: Assessing predictive validity.

Journal of Consulting and Clinical Psychology, 63, 737−748.Robertson, M. G., & Walter, G. (2008). Many faces of the dual-role dilemma in psy-

chiatric ethics. Australian and New Zeeland Journal of Psychiatry, 42, 228−235.Sandin, P. (1999). Better safe than sorry: Applying philosophical methods to the debate on

risk and the precautionary principle. Stockholm: Royal Institute of Technology.SBU (Statens beredning för medicinsk utvärdering) (Swedish Council on Technology

Assessment in Health Care (2005). Riskbedömningar inom psykiatrin: Kan våld isamhället förutsägas? En systematisk litteraturgenomgång. (Risk assessments inpsychiatry. Can violence in society be predicted? A systematic literature review)Stockholm: Statens beredning för medicinsk utvärdering SBU.

SFS 1979:680. Svensk Författningssamling. (Swedish Code of Statutes) (1979). Law onchange in the Penal Code. Department of Justice.

SFS 1991:1129. Svensk Författningssamling. (Swedish Code of Statutes) (1991). Law onforensic psychiatric care. Department of Justice.

SFS 1991:1137. Svensk Författningssamling. (Swedish Code of Statutes) (1991). Law onforensic psychiatric investigation. Department of Justice.

SFS 2006:45. Svensk Författningssamling. (Swedish Code of Statutes) (2006). Law onconversion of life-time prison terms. Department of Justice.

Soderstrom, H., Sjodin, A. K., Carlstedt, A., & Forsman, A. (2004). Adult psychopathicpersonality with childhood-onset hyperactivity and conduct disorder: A centralproblem constellation in forensic psychiatry. Psychiatry Research, 121, 271−280.

SOSFS 1996:14. Socialstyrelsens Författningssamling. (The National Board of Healthand Welfare Code of Statutes) (1996). Rättspsykiatrisk undersökning. (Forensicpsychiatric investigation) Stockholm: National Board of Health and Welfare.

SOSFS 2006:9. Socialstyrelsens Författningssamling. (The National Board of Health andWelfare Code of Statutes) (2007). Socialstyrelsens föreskrifter om säkerhet vidsjukvårdsinrättningar som ger psykiatrisk tvångsvård och rättspsykiatrisk vårdsamt vid enheter för rättspsykiatrisk undersökning. (Directives for safety infacilities providing compulsory psychiatric and forensic psychiatric care and inunits for forensic psychiatric investigation) Stockholm: National Board of Healthand Welfare.

SOFS 2006:16. Socialstyrelsens Författningssamling. (The National Board of Health andWelfare Code of Statutes) (2006). Riskutredning inför prövning av omvandling avfängelse på livstid. (Risk assessment in connection with evaluation of possibleconversion of life-time prison sentences.) Stockholm: National Board of Health andWelfare.

Steadman, H. J., & Cocozza, J. J. (1974). Careers of the criminally insane: Excessive socialcontrol of deviance. Lexington: Lexington Books.

Steadman, H. J. (2000). From dangerousness to risk assessment of community violence:Taking stock at the turn of the century. Journal of the American Academy ofPsychiatry and Law, 28, 265−271.

Tännsjö, T. (1999). Coercive care: The ethics of choice in health and medicine. London:Routledge.

Thornberry, T. P., & Jacoby, J. E. (1979). The criminally insane. A follow-up of mentally illoffenders. Chicago: University of Chicago Press.

Walters, G. D., Raymond, A. K., Grann, M., & Dahle, K. -P. (2008). Incremental validity ofthe Psychopathy Checklist Facet Scores: Predicting release outcome in six samples.Journal of Abnormal Psychology, 117, 396−405.

Webster, C. D., Douglas, K. S., Eaves, D., & Hart, S. D. (1997). HCR-20. Assessing the riskof violence. Version 2 Vancouver: Simon Fraser University.

Webster, C. D., & Eaves, D. (1995). The HCR-20 scheme. The assessment of dangerous-ness and risk British Columbia: Simon Fraser University and Forensic PsychiatricServices Commission of British Columbia.

Zweig, M. H., & Campbell, G. (1993). Receiver-operating characteristic (ROC) plots: Afundamental evaluation tool in clinical medicine. Clinical Chemistry, 39, 561−577.