Clinicians' Perceptions of Which Factors Increase or Decrease the Risk of Violence Among Forensic Out-patients

Please download to get full document.

View again

of 14
35 views
PDF
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Document Description
Clinicians' Perceptions of Which Factors Increase or Decrease the Risk of Violence Among Forensic Out-patients
Document Share
Document Tags
Document Transcript
  International Journal of Forensic Mental Health  2004, Vol. 3, No. 1, pages 23-36©2004 International Association of Forensic Mental Health Services Clinicians’ Perceptions of Which Factors Increase orDecrease the Risk of Violence Among Forensic Out-patients Knut Sturidsson, Ulrika Haggård-Grann, Malin Lotterberg, Mats Dernevik, and Martin Grann There is a debate in the academic community about which approach to the assessment of risk for violence is preferable. A vital aspect of this debate is evaluating the relative merits of clinical decision-making. However,clinical decision-making in violence risk assessment is understudied. The present study investigated if and how clinicians perceived that the absence and presence of dynamic factors affected the individual-specificrisk for violence in forensic out-patients. The Structured Outcome Assessment and Community Risk Monitoring (SORM) instrument was administered by 35 clinicians to 51 different patients, yielding a total of 103 uniqueclinician-patient encounters. The clinicians’ emphasized the following factors: lack of insight, lack of treatment motivation, psychiatric institutional treatment, professional support contacts, and substance misuse. Least weight was given to physical health care, children, occupational training and employment services, partner,and impaired daily functioning. The results suggest that clinicians considered violence risk and protective factors anchored in empirical research. Clinicians preferred clinical factors before non-clinical factors. Also, clinicians put more weight on individual than on contextual factors and more weight on risk factorsthan on protective factors. Knut Sturidsson, Ulrika Haggård-Grann, Malin Lotterberg, Mats Dernevik, and Martin Grann are all at the Centre for ViolencePrevention, Karolinska Institutet, P.O. Box 23000, SE-104 35 Stockholm, Sweden (academic); Knut Sturidsson and Martin Grannare also at the Regional Forensic Psychiatric Centre in Säter (clinic); Malin Lotterberg is also at Karsudden Forensic Hospital inKatrineholm (clinic); Ulrika Haggård-Grann is also at the Department of Forensic Psychiatry at the National Board of ForensicMedicine in Stockholm (clinic); and Mats Dernevik is also at the Regional Forensic Psychiatric Centre in Vadstena (clinic).Dr. Grann received financial support for this study from Vårdalsstiftelsen, The National Board of Forensic Medicine, and TheSöderström-Königska Foundation. We also thankfully acknowledge the support given by the Regional Forensic Psychiatric Centre,Landstinget Dalarna (Dalarna County Council).Correspondence concerning this article should be addressed to Knut Sturidsson, Karolinska Institutet, Centre for ViolencePrevention, P.O. Box 23000, SE-104 35 Stockholm, Sweden (Email: Knut.Sturidsson@cvp.se). The task of evaluating the risk that an incar-cerated offender or forensic patient will pose a danger to other members of society in the future and henceof deciding whether or not to grant parole or short-term community visits or to discontinue institution- based treatment is considered one of the mostdifficult and delicate tasks that clinicians and legal practitioners face. There is a debate in the academiccommunity about which approach to the assessmentof risk for violence is preferable (Litwack, 2001).One main point of this debate is whether and, if so,to what extent clinical judgment is useful for assessing the risk of violence. The proponents of actuarial risk assessment   have argued that “the placefor clinical judgment in violence risk assessment isin the gathering of relevant historical information,and careful scoring of actuarial instruments, rather than in replacing or adjusting the actuarially derivedscore” (Harris, Rice, & Cormier, 2002, pp. 391–392;see also Dawes, Faust, & Meehl, 1989; Grove &Meehl, 1996). These authors conceptually define risk assessment as the establishment of a probability of violence along a continuum (Quinsey, Harris, Rice,& Cormier, 1998). The approach emerged in the1980s as a reaction to the  Baxtrom  and  Dixon  cases(Steadman & Cocozza, 1974; Thornberry & Jacoby,1979). Those in favor of actuarial methods haveargued that it has been demonstrated that actuarialmethods outperform clinical judgment when it comesto risk prediction, and that this alone is an argumentfor the use of actuarial methods (Grove, Zald, Lebow,Snitz, & Nelson, 2000). Moreover, the assessment procedure should be transparent and the criteriashould be made explicit and available for scrutinyin court. Therefore, the assessment should be madein a statistical, mechanical fashion and solely on the  24Sturidsson, Haggård-Grann, Lotterberg, Dernevik, & Grann  basis of factors that have been demonstrated to bestatistically associated with violence. Clinical judgment should not be used.The actuarial approach has been heavilycriticized in recent years. Hart (2002, September)argued that it is for several reasons “impossible tounderstand, measure, predict, or communicate aboutthe behavior of individuals with precision.” Thecritics of the actuarial approach have argued that therisk factors and causal mechanisms of violence areyet poorly understood and that violence is not auniform phenomenon or a clearly defined outcome.Risk factors and outcomes materialize in a complexinteraction between human behavior and environ-ment. This complexity will not be appropriatelyaccounted for by a simple statistical algorithm.(Grann, 2002; Webster & Hucker, 2003). Critics havealso argued that actuarial instruments replicate poorlyacross different settings. Recent replication studieshave reported at best moderate ability for actuarial procedures to distinguish recidivists from non-recidivists (Barbaree, Seto, Langton, & Peacock,2001; Grann, Belfrage, & Tengström, 2000; Sjöstedt& Långström, 2001), although there is continuingcontroversy concerning the appropriate interpretationof data (Harris et al., 2002). Some of the actuarialresearch has been criticized for over-optimisticinterpretations of predictive validity estimates(Sjöstedt & Grann, 2002).Proponents for the use of  structured clinical  judgments maintain that evaluation of risk of futureviolence should be anchored in empirical researchand that clinical check-lists with factors predefined by previous research should be utilized. In addition,however, proponents for the use of    structured clinical judgments have argued that clinical judgment shouldsupplement and override any actuarial formulae(Hart, 1998; Litwack, 2001; Webster, Douglas,Eaves, & Hart, 1997). These authors do not claimthat the addition of clinical judgment totallyovercomes the obstacles described above. Rather, theadvocates for structured clinical judgment assert thatsimple algorithms can never replace the experienceand unique knowledge of the clinician about theindividual patient.A number of studies comparing actuarial andclinical decision-making have concluded thatactuarial predictions typically outperform unstruc-tured clinical predictions (Dawes et al., 1989; Groveet al., 2000). However, it has been argued that manyof the studies used to reach this conclusion have usednon-ecological methodology to the disadvantage of the clinician. The studies rarely use naturalisticsettings and do not allow the clinician to use his or her favored techniques for assessment (Rock,Bransford, & Mainsto, 1987). We agree with Litwack (2001), who noted that comparing the statistical predictive validity of an actuarial procedure withactual clinical decision-making on the risk of violentre-offending is neither just nor appropriate. Theexplanation is that clinical assessments are not predictions of future instances of violence, but rather an overview of potential risk factors pertaining tothe specific individual and considerations about howto decrease risk and prevent violence. In practice,“clinicians are bound—morally, ethically, andlegally—to prove themselves wrong when they‘predict’ violence; they must take every reasonableaction to ensure that those at high risk for violencedo not act violently” (Hart, 1998, p. 123).In essence it appears that a critical point in thecurrent controversy between actuarial versusstructured clinical judgment approaches to risk assessment is the relative merit of clinical decision-making. However, clinical decision-making in thisarea is understudied. With the distinction used byElbogen (2002), there are numerous  prescriptive studies (i.e. on which factors should be consideredfor violence risk assessment) but a paucity of  descriptive  studies (i.e. on what it is cliniciansactually do consider when making decisions aboutviolence risk). To evaluate the merits of clinical judgment, it is necessary to examine more closelyhow clinicians make decisions in ecological real-lifesituations. Background In agreement with Litwack (2001) and others,we argue that—with the current body of research— it is still difficult to make meaningful comparisonsof clinical and actuarial methods for at least threereasons. Firstly, the validity of clinical judgment hasusually been assessed via the use of vignettes rather than actual cases (for an overview see: Grove et al.,2000). Even when using actual cases as vignettes,the body of information typically consists of one or a few pages of written information as opposed to  Clinicians’ Perceptions of Risk25 real-life risk assessment, where the sources of information include files and reports, first-handknowledge of the patient via interviews, andexperience from the interaction between the assessor and the assessed. The generalizability of theconclusions drawn from experimental situations may be limited and the applicability of such conclusionto ecological assessments of violence risk may be poor.Secondly, a predominant approach to the studyof clinical decision-making is to ask clinicians torate the importance of a number of different risk factors in a hypothesized case. A specific exampleof this was a recently published study on the perceived relevance of various factors for violencerisk assessment (Elbogen, Mercado, Scalora, &Tomkins, 2002). In this study, clinicians were askedto rate the relevance of risk factors listed in aquestionnaire. Clinicians were not asked to consider any particular patient—and no clinical case vignettewas presented—but half of them were asked toconsider an admission context and half a dischargecontext. Elbogen and colleagues found that clinicians perceived dynamic behavioral variables (e.g., physical aggression whilst in care, impulsive behavior, medication non-compliance) to besignificantly more relevant than static-historicalfactors (e.g., childhood maladjustment, educationalhistory, marital history). However, these factors werechosen by the researchers in advance and may notnecessarily be the same ones the clinician would haveconsidered had he or she assessed the patient in anecological setting. Additionally, in any survey typeof study, factors will be presented to the informantsin a given order which may or may not be the sameorder they would have considered the factorsthemselves, and this may affect response style.Thirdly, the focus in most of the previous studieson the predictive validity of clinical judgment has been on risk factors rather than on protective factors(Rogers, 2000). Moreover, the interest has mainly been on individual factors (i.e. substance use, psychiatric symptoms) rather than contextual (i.e.housing, employment, family) factors. Thesedistinctions are not typical of clinical decision-making in an ecological setting. It may very well bethat clinicians utilize protective and contextualfactors as well as risk increasing and clinical factorsin the process of risk appraisal. If experimentalresearch is to examine the decision-making of theaverage or “generic” clinician when assessing risk of violence, the design must seek to better accountfor the use of risk vs. protective factors as well asthe absence vs. the presence of these factors.In the present study we argue that there is a lack of systematic investigation of which factorsclinicians actually consider in the assessment process. Furthermore, the clinician typically has awide range of responsibilities vis-à-vis the patient,the mental health service, and the community;assessing and managing the risk of future violenceis but one. As opposed to the researcher—whoutilizes a set of operationalized risk factors on a  group of forensic psychiatric patients to assess the predictive validity in a follow-up study—theclinician is expected to do the same in individual  cases and to act upon the observations made, withfar-reaching consequences for the patients and othersif decisions prove incorrect. Consequently, clinicaldecision-making should preferably be studied under these ecological conditions.The present study addressed one facet of the risk assessment process, namely, the clinicians’ percep-tion of which factors increase or decrease the risk for violence among forensic out-patients. In addition,we examined how the clinicians motivated specificfactors such as risk or protective factors. To the bestof our knowledge, there are no published studies onthis subject in a naturalistic setting in which cliniciansassess patients in the here-and-now and not vignettesin retrospect.We acknowledge that the present study will notfurther any comparison of clinical versus actuarialrisk assessment but rather contribute more know-ledge to how clinicians perceive risk factors. METHODSetting The present study was based on data from theCOMET project (which is an acronym for contextualfactors that mediate violence risk), a multi-centrefollow-up study of former forensic psychiatric patients, which is described in further detailelsewhere (Grann et al., in press). The data in the present study were drawn from the second and largest  26Sturidsson, Haggård-Grann, Lotterberg, Dernevik, & Grann  part of the project. This part (II) is a prospectivefollow-up study where forensic patients are followedonce a month for two years during their transitional period from release from an institution via mandatedout-patient management into society. The other two parts of the COMET project are a case-control study(I) in which patients (cases) who recidivated withinfive years of a prior offence are compared with patients (controls) who did not recidivate and a case-crossover study (III) of trigger factors for violence.The study population in the present study wasdrawn from individuals who have been found guiltyof violent crime and judged to suffer from a severemental disorder in a forensic psychiatric evaluation.Under present Swedish legislation, offenders whoare judged to suffer from a severe mental disorder will be diverted to a forensic hospital instead of beingimprisoned (Grann & Holmberg, 1999). Nine different forensic psychiatric facilities inSweden are involved in the project. Hereinafter theforensic patients will be referred to as informants ,and the clinicians will be referred to as raters . Ethicalapproval for the study was granted by the regionalresearch ethics committee of Karolinska Institutet(ref. # 99-416). The SORM The raters administered the Structured OutcomeAssessment and Community Risk Monitoringinstrument (SORM; Grann et al., 2000). The SORMis designed to be a clinical tool for continuousmonitoring of the risk for violence posed by forensicmental health clients in the community. The factorsassessed can be seen in Figure 1. The rating procedure in the SORM consists of two parts: one part assesses the absence or presence of the differentfactors defined by the manual, the other concernsthe effect these factors may have on the risk level. If the circumstance described in the item is deemed to be absent, the item is rated “No.” If a circumstanceis present, it is rated accordingly with “A”, “B”, or “C”, which denotes different degrees and types of  presence of the factor. For instance, if SORM #2 is positive—psychiatric treatment is considered present—the rater must specify whether the presencereassembles “the individual has been registered as a patient on a psychiatric ward throughout the monthand has spent practically all that time on the ward”which equals an “A” or “the individual has beenregistered as a patient on a psychiatric wardthroughout the month, but has often spent timeoutside on leave, or has only been committed part of the month” which renders a “B” or “the individualhas been registered as a patient on a psychiatric ward, but has only made short visits to collect medication,go to therapy, take part in social events, etc.” whichdefines a “C” (Grann et al., 2000, p. 7).This part of the rating procedure strives for asmuch objectivity as possible, and the operationaliza-tion of factors is “actuarial” in the sense that thereare strict, predefined criteria for the presence or absence of factors.The first five items of the SORM cover aspectsof current services and interventions provided byagencies and services in the community. The second block of five items covers aspects of the informant’s personal circumstances, called “social situation.” Afurther four items concerns the informant’s socialnetwork, which is defined as “the people around the person who do not have a professional relationshipwith him or her” in the SORM manual (Grann et al.,2000, p. 26). Ten clinical items focus on mentalhealth problems. And finally, three self-rated itemsare included where the informant indicates his or her subjective rating of health, quality of life, and risk of recidivism. The SORM also includes threeoutcome variables (# 28–30) where the rater usesthe informant’s self-report as well as agency recordsto record occurrences of violent behavior, exposureto high risk situations for violence, and other criminal behaviors in the period since the previous SORMassessment.In an inter-rater reliability study of 20 cases ratedusing SORM, the reliability of the factors asmeasured by calculating Cohen’s Kappa (Cohen,1960) was on average = 0.88 (range 0.32–1.00,median = 1, mode = 1), which is considered excellentaccording to prevailing statistical interpretationstandards (Altman, 1991; Cicchetti & Sparrow, 1981). The three items that had the lowest inter-rater reliability were SORM #22, lack of treatmentmotivation (0.32); SORM #17, anxiety (0.44); andSORM #15, lack of insight (0.50). All other itemshad Kappa values corresponding to “good” or “excellent” reliability according to Cicchetti andSparrow.  Clinicians’ Perceptions of Risk27 The second part of the assessment procedure inthe SORM concerns the risk effect. The rater isinstructed to judge how the risk for violence isinfluenced by the absence or presence of each of thefactors. This is a subjective rating made by the ratingclinician who, in this part, is encouraged to use hisor her clinical judgment through clinical knowledge, professional experience, and “gut feeling”. Theclinicians rate whether they judge the circumstanceat hand as increasing, decreasing, or not affectingthe risk of violence. This is done by assigning theitem a “-“ (minus), a “+” (plus), or a “0” (zero),respectively. The presence or absence of any givencondition may act as a risk factor for one individual but a protective factor for another individual. It isthe rating of risk effect that is the focus of the presentstudy.In addition to the two parts described above, theSORM instrument also comprises an idiographical  part  . The rating clinician describes in free text thecircumstances that apply to each factor in eachindividual case. For example, item 23—pharma-cological treatment—may have notes on which drugsand doses the forensic mental health client has been prescribed and item 6—housing—may include noteson more exact circumstances on how the client lives(e.g., “is living together with his mom in X-town, ina 4-bedroom apartment”). When the SORM is usedclinically to monitor forensic mental health clientsin the community at regular intervals, the clinicianis instructed by the manual to make notes in theidiographical section when and if changes in thefactors occur.The assessment of risk effect also contains anidiographical part. The rating clinician is hereinstructed to motivate how and why he or she hasrated a particular factor as increasing or decreasingthe risk. For example, if the patient living with hismom was rated “+” on item 6, the idiographicalsection may read “because he and his mom haverecurrent conflicts over household issues, and theseconflicts have in the past led to violent incidents”. Participants The data set in this study was drawn frominterviews between raters and informants in the on-going, prospective follow-up study. Data from theinterviews were coded in the format described bythe SORM manual (Grann et al., 2000).The raters ( n = 35) were clinical psychologists(37.1% of the interviewers, 45.6% of the interviews), psychiatric nursing staff (57.1% of the interviewers,48.5% of the interviews), and social workers (5.7%of the interviewers, 5.8% of the interviews). Amongthe raters there were 19 (54.3%) female raters and16 (45.7%) male raters. Their mean age was 45.43( SD  = 9.39) years. Mean clinical experience inforensic psychiatry was 11.09 years ( SD  = 8.17) andranged from 1 year to 30 years.All of the informants ( n = 51) were forensic psychiatric patients in various stages in the processof being released into the community (spending atleast 50% of their time outside the forensic psychiatric institution). Among the informants were42 (82.4%) males and 9 (17.6%) females. Their meanage was 40.59 ( SD  = 11.20) years. All of theinformants were violent offenders who had beendiverted to forensic psychiatric care on account of mental disorder, and they had all been assessed to be at high risk of serious criminal recidivism.The SORM interviews and protocols wereadministered by 35 clinicians to 51 different patients,yielding a total of 103 unique clinician-patientencounters. Fourteen raters assessed only oneinformant each, 8 raters assessed two informantseach, and 13 raters assessed more than twoinformants each. The units of observation in the present study are thus these 103 unique constellationsof a rater and an informant, which henceforth arereferred to as cases. Measure The SORM variables were ordered by rank. Therank order was established by analyzing which of the variables were perceived to exert a risk effect (ineither direction [+] or [-]) in most of the cases. Anytype of presence, that is A or B or C coding, wascollapsed into and then analyzed as a categorical“any” presence of factor (the left-hand column inthe coding sheet, see Figure 1). A measure of therelative frequency of observations in which the raters’ perception of the factor increased or   decreased therisk was calculated. The criteria variables (violence,other criminal acts, and risk situations) were notstudied in the present paper.
Similar documents
View more...
Search Related
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks