Week 7.2 Assessment of Risk and Dangerousness 2024 PDF

Assessment of risk and dangerousness in Forensic Psychology Howitt Chapter 25  Prediction Psychologists have many responsibilities when undertaking forensic work including: – Protection of general public from...

Assessment of risk and dangerousness in Forensic Psychology Howitt Chapter 25  Prediction Psychologists have many responsibilities when undertaking forensic work including: – Protection of general public from dangerous individuals – Protection of individuals from self-harm and suicide – Protection of staff of institutions – Protection of other inmates These require the psychologist to make predictions. Seek to predict different things – e.g. risk of offending or risk of offending in a particular way Predicting Risk is about predicting likelihood of offending Predicting Dangerousness is about predicting likely consequences of offending – how “serious” the offence. All combinations of risk and dangerousness levels are possible – for example can predict high risk but low dangerousness.  Tarasoff Decision Californian court case in 1970s – Tatiana Tarasoff was a student at U California in late 60s- early 70s – Met another student Prosenjit Poddar. Breifly dated but Tarasoff ended relationship. Poddar started to stalk Tarasoff. Mental health declined. – Poddar told University health staff that intended to kill Tarasoff. Psychologist Dr Moore asked campus police to detain Poddar. – Poddar was briefly detained but then released after Moore’s boss ruled he was not to be subject to detention. – Tatiana was not told of threat – Poddar killed Tatiana – Court ruled that mental health professionals are legally obligated by duty to inform such potential victims of threat to their safety made by clients – Revised in 1976 - must use reasonable care to protect potential victims (not necessarily by informing them directly) – So how do psycs assess risk and dangerousness?  Clinical Vs Statistical assessment 2 types of risk and dangerousness assessment – Clinical judgment – based on experience of clinician. – Statistical or Actuarial assessment – based on evidence from previous cases which identity risk factors Evidence is that Actuarial is more accurate  Clinical approaches Several models of clinical decision making re risk; – Linear model – series of simple decisions eg- Is there a clear (not vague) threat? Serious (not marginal) danger? Specific victim (not non-identifiable) ? Imminent danger? – Hypothetico-deductive model. Knowledge about previous behaviour allows clinician to formulate hypotheses about likely future behaviour. See example in Howitt of young man threatening violence to girlfriend. Counselor knows student has history of alcohol abuse and fighting in school. Father is overseas, mother reports cannot control her son.  Clinical assessment of risk Many studies show clinical assessments of risk to be poor Clark (1999) reviewed studies and concluded that clinical risk assessment is weak at best, at worst totally ineffective. Even experienced clinicians fail to predict future violence in cases with clear indicators, such as previous recidivism.  Why do clinical judgments get it wrong? Dernevik (2000) suggest; Diagnostic categories are too broad (eg schizophrenia) so allocating someone to such a category is not helpful in predicting future behaviour Limited capacity of human information processing systems means we may pay too much attention to just a few factors and ignore others (heuristics) Biases – we tend to stick to initial judgment and often ignore new information that is inconsistent with this. Also we have a tendency to identify illusory correlations - things we spot in a few cases and then only look for confirming evidence There can be too much information to deal with Experience doesn't necessarily help because we often don’t get full feedback. For example if a clinician predicts someone will reoffend they may be held in custody – so never know if was correct or not.  Improving Clinical methods Blackburn (2000) – provision of guidelines to help structure clinical decision making can improve performance Hollin and Palmer (1995) point out that the superior performance of actuarial methods suggests that individual differences or environmental stressors are not important in determining behavior – this seems very unlikely. The problem is that it is difficult to incorporate these factors into risk assessment in a reliable manner. Should distinguish between clinical variables (should be included) and clinical judgment (unreliable) PCL-R includes some clinical variables such as “superficial charm”  Statistical or Actuarial prediction Empirically based method Calculate risk by comparing characteristics of the individual in question to other individuals for whom we know the outcome Originates from work of insurance industry who want to calculate such things as risk of car crash or risk of dying in next 10 years. Actuaries produce “life tables” (“An actuary is like an accountant, but with fewer social skills!”) Draws on databases of characteristics of offenders and offence histories Based on factors like age, gender, age at first offence, personality Examples of Life Tables Some more examples! How reliable are individual predictions?  The accuracy of Prediction Reoffends Doesn’t Reoffend Shows predictive True Positive False Positive characteristics Doesn’t show predictive False Negative True Negative characteristics We want to maximise True Positives and True Negatives, but minimise False Negatives and False Positives. Note that these errors may not be equally important The accuracy with which we can predict an event is related to the base rate frequency of that event – our best predictions occur when the behaviour occurs at a frequency of 50%.. So, if we are dealing with an infrequent event – such as violent offending it is very difficult to accurately predict offending and many of our predictions will be wrong It is VERY hard to predict some offences  The accuracy of predictions The question of the accuracy of predictions of dangerousness is an issue sometimes examined by NSW courts which have to consider whether to detain someone beyond the end of their sentence if thought to pose a very serious risk (preventative detention) Continuing Detention Orders for Serious Sex offenders More recent introduction of similar legislation for violent offenders  An example – Hollin & Palmer (2001) Sought to predict reconviction. Only significant predictors were age and criminal history. On basis of these can make predictions – see table Overall accuracy was 72% ((23+130)/(23+18+41+130)), but this hides a more complex picture. Only slightly more than half of the predictions of reconviction were accurate (23/(23+18))= 56%. 76% of predictions of non-reconviction were accurate (130/(130+41)). Of all the people reconvicted we only predicted correctly for 36% (23/(23+41)). Of all the people Not reconvicted we predicted correctly for 88% (130/(18+130))  Predictors The actuarial approach identifies predictors – factors which significantly predict future offending – either on their own or in combination E.g. Gretenkord (2000) sought to predict violent recidivism in mentally disordered offenders. Analysed data and identified just 4 predictors: – Presence of personality disorder (Y/N) – Was there a violent offence prior to offence that led to institutionalisation (Y/N) – Physical aggression while in hospital (Y/N) – Age at time of discharge Young offenders (20 years) with a personality disorder, violent pre-offence who had been aggressive during treatment showed a 65% chance of violent reoffending. This drops to just 16% for similar offenders aged 60 years. Just 1% if no personality disorder, no violent pre-offence and no aggression during treatment for a 60 year old. Note that some of these factors are clinical in nature, but the approach is actuarial  Best predictors The predictors of future offending vary with offence type (eg domestic violence predicts future domestic violence better than it does other sorts of violent offences) Some common predictors appear in many prediction instruments These include age, criminal history, substance abuse  Static and Dynamic factors Static factors cannot be modified – eg age at first offence Dynamic factors can change – eg martial status Many predictive factors are static. This creates some problems – does it mean we cannot change dangerousness But note that predicative influence of a static factor can change over time – for example as the offender gets older the risk predicted by an early offence might decline Also, what about personality factors, are they static or dynamic?  Other factor types Blackburn (2000) distinguishes between: – Historical factors (especially past offending) – Dispositional factors – cognitive and emotional tendencies – Clinical variables – mental disorders – Personality factors – personality disorders  Latest prediction instruments The latest generation of prediction instruments are designed to not only predict risk, but also to identify risk factors and to help develop risk management plans (eg HCR-20) Some of these instruments allow the clinician to modify assessed risk level on basis of clinical judgments – called Structured Professional Judgment (SPJ) model Some include protective factors (eg family support) which might lower risk  The most difficult cases There is no evidence that we can predict violence in individuals who have no history of violent offending (eg high school shooters etc) Some people would argue that this is one of the most important targets of prediction  Conclusions Prediction of risk and dangerousness is an important part of the role of forensic psychologists We are getting better at it but many instruments are not as accurate as sometimes portrayed Much more work needed