Risk Assessment in Forensic Learning Disability Psychology

1
Risk Assessment in Forensic Learning Disability
Psychology

• Dr Jeremy Tudway
• Consultant Forensic Clinical Psychologist/Lecturer
  -Practitioner in Clinical Psychology

2
Risk Assessment Context

• Justice is now very much less important than
  risk, as a preoccupation of criminal justice/law
  and order policy the politics of safety have
  overwhelmed attachment to justice in the
  institutions of late-modern demographic polities
• Gray, et.al., (2001)

3
What should risk assessments do? i

• Identify likely outcome
• Estimate
• Likelihood
• Probability
• Identify factors
• Increase
• Decrease
• Reliable and valid

4
What should risk assessments do? ii

• Probability NOT Possibility
• The LIKELIHOOD of a given event
• Not the HYPOTHETICAL chance
• It is possible that anyone could set a fire,
  whereas the probability of someone with no
  fire-setting history, no interest in fires and
  few interpersonal problems deliberately setting
  fires is small in comparison to someone who has
  previously set fires, shows a long-term interest
  and has gross interpersonal problems

5
Problems with Risk Assessment

• Hired guns
• Ill-informed appraisal
• Poorly researched instruments
• Inadequate base-rate estimates
• Not really knowing the true rate of occurrence
• Actuarial versus Clinical data
• But, crime detection and conviction does not
  necessarily equate to the true rate at which
  offending has taken place.

6
Clinical versus Actuarial i

• Clinician accuracy is low (violent behaviour).
• Monahan (1981)
• Clinical
• Not systematic or standardized, based on
  judgment.
• Evidence suggests that professional judgement
  is generally very poor and fraught with biases
• Mental health professionals often make critical
  decisions in the first couple of seconds of a
  contact and then direct questions towards
  proving their assumptions!
• Actuarial
• Systematic, standardized, rule-governed.
• Empirical literature
• Clinical tends to be more conservative.

7
Clinical versus Actuarial ii

• All risk assessment procedures still require
  professional judgment
• Choice of scale, or critical variables
• Interpretation of scores
• Using idiosyncratic or clinical judgement
• No systematic empirical support
• Low agreement (unreliable) and accuracy
  (unvalidated)
• Actuarial risk assessment by trained assistants
  has stronger predictive validity than therapists
  assessment.
• Grove Meehl (1996)
• 40 of better, 40 no difference, 10 poorer
• Beech Ward (2003)

8
Six risk assessment methods(Doren, 2002)

• Unstructured/unguided clinical judgment
• Case material without any structured assessment.
  Decisions are intuitive or experiential.
•  Purely actuarial approach
• Straightforward algorithmic procedures and
  risk-prediction instruments. Reliant on
  historical/static risk factors and coding to
  arrive at a probability of reconviction.
•  Guided clinical judgment
• Clinical judgment without relating to current
  theories of risk.
•  Anamnestic risk assessment
• Life history examined in relation to
  dispositional contextual factors. Current
  circumstances examined for presence of particular
  identified risk factors
•  Research-guided clinical judgment.
• Apriori set of research-informed factors (e.g.,
  SVR-20) as a guide to assessment.
•  Clinically adjusted actuarial approach.
• Initial employment of one or more actuarial
  instruments is followed by potential adjustments
  based on clinical/dispositional considerations.

9
What is a predictor?

• it is often said that the best predictor of
  future violence is past violence, and anyone with
  a history of violence needs to be managed with
  care..
• Unless there are good reasons to believe that
  major changes have occurred, it is always wise to
  assume that violence is a possibility, and to
  take steps accordingly
• Sellars, (2002)

10
In other words.
A TINY bit more work required here I think?
11
Structured Risk Assessment

• Many different devices
• SACJ (Thornton)
• RM2000 (Thornton)
• VRAG (Quinsey, Harris, Rice Cormier)
• HCR-20 (Webster)
• RRASOR (Hanson)
• SVR-20 (Boer, Hart, Kropp Webster)
• OASYS (Home Office)
• Approximately 26 specific sex offender risk
  assessments to date! (Doren, 2002)

12
Static Variables

• Tombstone factors
• Always additive
• Usually negative
• Although the long-term probability may reduce
  evidence is unclear
• Tend to out-weigh other sources of data
• (eg PCL-r)

13
Dynamic Variables

• Indicators of change
• Dynamic
• Can reduce and increase
• Can be both positive and negative
• Linked to harm-reduction or relapse-prevention
  models
• Can be linked to intervention
• Ideally lend themselves to psychometric data
• (eg SOAP)

14
Acute Stable Dynamic Variables

• Acute dynamic factors
• Indicate high likelihood of an offense in the
  near future. (Hanson Harris, 2001)
•  Integrated model that includes
• Distal factors (developmental)
• Historical and stable dynamic (vulnerability or
  trait)
• Triggering events (contextual factors) that
  combine with static and stable dynamic factors
  and this drive
• Acute dynamic (state) factors.
• (Beech Ward, 2003)

15
Base Rates

• The frequency of the target behaviour in any
  given population
• Accurately know the prevalence
• Offending behaviour is grossly underestimated by
  conviction
• The rate at which acts occur in the population of
  interest is critical to the predictive ability of
  any instrument.

16
The Base Rate Fallacy

• Failure to take base rates into account when
  judging probability.
• When considering a particular case, the
  statistics for that population should also be
  considered.
• People tend to ignore the population base rate
• People are not good statisticians!

17
Problems with Prediction i
Hits
Hits
True Positive
True Negative
Misses
Misses
False Positive
False Negative
18
Problems with Prediction ii

• False positives
• Ethical
• Restriction and impingement of human rights
• Creating more problems (vicarious learning,
  institutionalisation)
• Alarm and concern to systems
• Cost
• False negatives
• Ethical
• Exposing people to unnecessary potential harm
• Undermining public confidence in services
• Damage to public perceptions of PwLD
• Cost

19
Problems with Prediction iii

• The most widely research tools report good
  sensitivity using the Receiver Operator
  Characteristic (ROC) curve
• Estimates the accuracy of the instrument by
  calculating how much data can be accounted for
  using
• Sensitivity
• True positive rate
• Hits
• Specificity
• True negative rate
• Misses

20
Problems with Prediction iv

• No matter how good the ROC curve, with low
  frequency, high impact behaviour.
• Over estimation of risk and false positives
• Positive predictive value (the proportion of
  positive predictions that turn out correct)
• (eg Szmukler, 2001a,b)

21
Problems with Prediction v

• Statistical significance is not necessarily
  linked to clinical significance.
• The probability of a given number arising on an
  analysis may be remote but this does not suggest
  that this figure is of particular significance to
  an individual

22
New ways of thinking?

• Combining both Actuarial Clinical data
• (Beech Ward, 2003)
• Dynamic Static factors
• Situational Accidental triggers
• The dichotomy between clinical and actuarial is
  unhelpful and being replaced by hybrid models,
  such as the HCR-20
• (Johnston, 2002)
• Despite this increasing sophistication of
  research in mainstream forensic psychiatry, the
  ability to predict future offending behaviour
  remains very limited.
• Turner (2000)

23
How appropriate are existing schemes?

• Can actuarial predictors developed in psychiatric
  or prison populations be valid for individuals
  with intellectual disabilities?
• (Green et.al., 2002 Johnston, 2002 Turner,
  2000)
• Adapted SACJ significantly inflates risk estimate
  in an LD group
• Under reporting and variable legal responses to
  offending behaviour in PwLD.
• Conviction not associated with
• Volume
• Victim range
• Risk status
• (Green et.al., 2002)

24
Challenges i

• Methodological problems
• Sampling (range, borderline)
• Definitions (Violence, Sexual offence)
• Convicted versus non-convicted (Cautions)
• (eg Green et.al., 2002 Lindsey, 2002)
• Rationale for good practice is ill-defined
• (eg Johnstone, 2002)

25
Challenges ii

• Difficult to make reasonable comparisons
• Reported re-offending rates vary from study to
  study 0-28 31.3-85
• (Barron et.al., 2002)
• Difficult to establish predictive factors
• Few differences on factors associated with sexual
  recidivism
• (Green et.al., 2002)
• Actuarial measures share a common theoretical
  heritage
• How appropriate to use multiple forms?

26
The Future i

• Co-ordinated approach to research and data
  gathering
• Initiate inclusive collaborative research to
  retrospectively identify key predictor variables
  using robust technologies
• Initiate a longitudinal prospective study using
  identified factors
• Develop culturally sensitive reference groups
• Ethnicity, Social expectations

27
The Future ii

• Develop meaningful normative models for comparing
  key behaviours
• Sexual (eg beliefs, behaviours), Interpersonal
  (eg assertion), Motivational (eg
  impulsivity), Social (eg gender roles) etc
• Include cautions and systematically detailed
  reports of problem behaviours in addition to
  convictions
• Develop functional formulation
• Develop Offence Parallel Behaviour models
• Make the technologies available
• Establish government funding
• Communicate the findings and change practice as a
  result!

28
A final point?

• Despite all the complications there are, however,
  two things one can predict with certainty

Death and Taxes
29
References

• Barron, P., Hassiotis A. Banes, J. (2002)
  Offenders with intellectual disability the size
  of the problem and therapeutic outcomes. Journal
  of Intellectual Disability Research. 46, 6, 454
  463.
• Beech, A.R. Ward T. (2003) The integration of
  etiology and risk in sexual offenders A
  theoretical framework. Aggression and Violent
  Behavior. 295 In Press
• Doren, D. M. (2002). Evaluating sex offenders A
  Manual for Civil Commitments and Beyond. London
  Sage.
• Gray, N.S., Laing, J. Noakes, L. (Eds) (2001)
  Criminal Justice, Mental Health and the Politics
  of Risk. London Cavendish Press.
• Green, G., Gray, N.S. Willner, P. (2002)
  Factors associated with criminal convictions for
  sexually inappropriate behaviour in men with
  learning disabilities. The Journal of Forensic
  Psychiatry. 13, 3, 578-607.
• Grove, W. M., Meehl, P. E. (1996). Comparative
  efficiency of informal (subjective,
  impressionistic) and formal (mechanical,
  algorithmic) prediction procedures The clinical
  statistical controversy. Psychology, Public
  Policy, and Law, 2, 293323.
• Hanson, R. K., Harris, A. (2001). The sex
  offender need assessment rating (SONAR) A method
  for measuring change in risk levels. Available
  www.sgc.gc.ca/epub/corr/e200001a/e200001b/e200001b
  .htm.
• Johnston, S. J. (2002) Risk assessment in
  offenders with intellectual disability the
  evidence base. Journal of Intellectual Disability
  Research. 46 Sup1, 47-56.
• Lindsay, W.R. (2002) Research and literature on
  sex offenders with intellectual and developmental
  disabilities. Journal of Intellectual Disability
  Research. 46 Sup1, 74-85.
• Monahan, J. (1981) Predicting Violent Behavior
  An Assessment of Clinical Techniques. Beverley
  Hills Sage.
• Sellars, C. (2002) Risk Assessment in People with
  Learning Disabilities. BPS Books, Blackwell.
• Szmukler, G. (2001a) The mathematics of risk
  assessment for serious violence. Psychiatric
  Bulletin 25 359
• Szmukler, G. (2001b)Violence risk prediction in
  practice. The British Journal of Psychiatry. 178
  84-85
• Turner, S. (2000) Forensic Risk Assessment in
  Intellectual Disabilities The Evidence Base and
  Current Practice in One English Region. Journal
  of Applied Research in Intellectual Disabilities.
  13, 4, 239