Sally C. Johnson MD
Professor
UNC Department of Psychiatry
Forensic Psychiatry Program and Clinic
Grand Rounds Presentation - 5/25/11
Dead Right,
Dead Wrong,
or the Jury is Still Out:
The Complex Worlds of Violence and Mental Illness
Learning Objectives



Appreciate the complexity of the
relationship between violence and
mental illness
Understand the current state of risk
assessment
Translate this understanding into a
practical approach to risk management
75% of people believe that
people with mental illness are
dangerous.
Literature:
Mental Illness and Violence
From Nursery Rhymes…
Lizzie Borden took an axe
and gave her mother forty whacks.
When she saw what she had done
she gave her father forty-one.
To recent best-sellers…
“These were the lovely bones that had grown around
my absence…”
Susie Salmon cutting through a
cornfield after school is persuaded
by George Harvey, a man in his
mid-40s who lives alone and builds
dollhouses for a living, to have a
look at his underground den. He
rapes and kills her, dismembers
her body, puts the parts in a safe and
dumps it into a sinkhole.
We are led to believe that “crazy” people
do crazy and frightening things.
Film Portrayal:
Mental Illness and Violence
The idea of the Insane Killer…
Takes a real life story like
that of Ed Gein
Norman Bates in
Alfred Hitchcock’s
“Psycho”
In the News:
Mental Illness and Violence
Whether it is the poor handyman in need of a job…
Bryan David Mitchell
Elizabeth Smart Kidnapping
Or workplace violence that hits close to home…
The murder of an NIMH administrator while trying to help a
psychotic patient sent shockwaves through the mental health
community, forcing clinicians to remember the rare—but ever
present—risk of violence. It is a rare scenario, the potential
nightmare in the life of a psychiatrist: a patient becomes
violent…while the psychiatrist and the patient are alone in the
psychiatrist's office.
Wayne Fenton, M.D.
October 3, 2006
…or a psychiatrist (or terrorist ?) turned mass
murderer…
Nidal Malik Hasan
Ft. Hood Killing Rampage
We are surrounded by possible links
between violence and mental illness,
and it’s frightening.
We look for ways to give
names and faces to our
fears…
…we want to know who is
going to be violent, and we
want to stop them before
violence strikes.
Cesare Lombroso
Violent criminals are:
• throwbacks to primitive humans
• identifiable by physical characteristics
• we can detain or execute them
Physical Signs of Imminent
Violence (Berg, Bell, and Tupin, 2000)








Chanting
Clenched Jaw
Flared Nostrils
Flushed face
Clenched or Gripping hands
Darting Eye Movements
Increased proximity of patient to Clinician
Inability of Patient to Comply with
reasonable Limit setting
Core Issues to Consider
Definitions
Violence / Mental Illness
Assessment
Adequacy / Frequency
Prediction
Of What / Duration
Prevention
By Whom / At What Cost
Responsibility
Liability / Blame
Violence: What Does It Mean?







Actual physical violence
Potential for violent behavior
Threat of violent behavior
Breaking the law
Psychological or emotional harm
Risk to property
A specific act or just a general propensity towards
violence
Violence: Dimensions
 Imminence
 Frequency
 Severity
 Setting
 Direction
What about defining
Mental Illness or
Mental Disorder?
Clinical Definitions


ICD-10 : “the existence of a clinically recognizable
set of symptoms or behavior associated … with …
interference with personal functions.”
DSM-IV-TR : “a clinically significant behavioral or
psychological syndrome or pattern that occurs in an
individual …associated with present distress …
disability … or a significantly increased risk of
suffering death, pain, disability, or an important
loss of freedom …

DSM-V: ???
BUT…
Within the legal/judicial system,
mental illness or disorder is
viewed as a legal, moral or
policy judgment or definitionnot a clinical one.
Our legal system has long connected
mental illness and violence and looked
to clinicians to predict likelihood of
future violence:
Legal Areas of Violence Risk
Prediction

Civil Commitment
O’Connor v. Donaldson (1975) / Addington v. Texas (1979)

Civil Liability
Tarasoff v. Regents of the University of CA (1976)
Death Penalty Cases
Jurek v.Texas (1976) / Barefoot v. Estelle (1983)
 Juveniles
Shall v Martin (1984)
 Preventive Detention
U.S. v. Salerno (1987)
 Sex Offender Commitment Statutes
Kansas v. Hendricks (1997) / US v Comstock (2010)

“
Legal dependence on clinical prediction
of risk has persisted despite data
suggesting that clinicians are often
wrong in their predictions.
Natural experiments
suggesting clinicians were
wrong more often than not:

Baxstrom v. Herald, (1966)-release or transfer of “dangerous”
patients. In 4 yr. follow-up only 20% assaultive

Dixon v. Attorney General of Commonwealth of PA
(1971/1979-review)-86% false positive rate among those originally predicted to be
dangerous


1976- Cocozza / Steadman -257 incompetent felons
released: 14% of those predicted as dangerous
were rearrested; 16% of those not viewed as
dangerous were rearrested!
1977-Patients released from Patuxent Institute/ MD
-58% false-positive rate in predictions of violence
In the wake of early legal decisions, research
efforts increased with the aim to improve
violence risk assessment in clinical practice and
the criminal justice context by:
– identifying empirically-validated risk factors
– developing risk assessment instruments
based on empirically-validated risk factors
Methodological problems in earlier
studies were identified …






Large #s of patients were lost to follow-up
Many had been treated for years
Reviews relied on official criminal records
which grossly under-estimated violence
Definition of violence was inconsistent
Original predictions had not all been clinical:
many were administrative or legal
Turns out that Clinicians were actually right
more often than not - but just barely
Review of History of Study of
Relationship between Violence and
Mental Illness




Pre-deinstitutionalization studies showed no increased
risk of violence
Post-deinstitutionalization studies began to show
increased risk
It appeared that increased risk might be more
connected to active symptoms rather than to diagnoses
There was more and more evidence that the
relationship between mental illness and violence was
actually quite complex
MacArthur Violence Risk
Assessment Study (1994)




Civil Admissions from inpatient psychiatric hospitals: Western
Psychiatric/ Pittsburgh, PA; Western Missouri Mental Health
Center/ Kansas City, MO; Worcester State hospital and
University of Massachusetts Medical Center / Worcester, MA
Ages 18-40
English Speaking / White or African-American (Hispanic at
Worcester)
Chart Dx of Schizophrenia, schizophreniform, schizoaffective,
depression, dysthymia, mania, brief reactive psychoses,
delusional disorder, alcohol or drug abuse or dependence, or
personality disorder. Research and clinician interviews in
hospital; two research interviews of patient and collateral
informant with next 20 weeks. Review of hospital, arrest and
rehospitalization records
MacArthur Study-18.7 % of patients were
involved in violent altercations:
Significant Findings






Men no more likely to be violent than women; drinking , SA
and medication non-compliance > in men / women directed violence
against family and at home
All measures (self report, hospital and arrest records)- previous
violence and criminality strongly related to future violence
Prior physical abuse, but not sexual abuse as child was associated
with post-DC violence
Parents history of substance abuse or criminal behavior:
strong relationship
All races in same disadvantaged neighborhood had same risk:
crime rate of neighborhoods pts. are discharged into may be
important factor
Personality disorder/ adjustment disorder had greater risk
than all other Dx; schizophrenia<depression or bipolar but > than
non-disordered population
MacArthur Study Findings
continued….






Co-occurring Dx of Substance Abuse or Dependence
strongly predictive
Psychopathy (the antisocial component) as measured by
PCL predicted violence
Delusions were not predictive (even threat-control-over-ride)
but suspiciousness was
Hallucinations/ command hallucinations were not predictive
unless voices specifically commanding violent acts
Persistent violent thoughts during hospitalization and
afterwards were predictive
Anger: high scores on Novaco Anger Scale at hospitalization
were twice as likely to engage in violent acts post DC
Where Are We Now?
National Epidemiologic Survey on
Alcohol and Related Conditions
(NESARC)
We employed a nationally representative,
longitudinal dataset from this two wave, face-toface survey conducted by the National Institute on
Alcohol Abuse and Alcoholism. N= 34,653 subjects
Wave 1 (2001-2002) Wave 2 (2004-2005).
Our questions were:
1) Does severe mental illness (SMI) predict future
violent behavior?
2) What risk factors prospectively predict violent
behavior?
Multivariate Predictors of Violent
Behavior Perpetrated Between
Waves 1 and2




Dispositional Factors: age, education, sex, race, income
Historical Factors: parental criminal history, witnessing
parental fighting, history of any violence, history of juvenile
detention
Clinical Factors: Schizophrenia, Bipolar Disorder, Major
Depression, Substance abuse/Dependence,
Schizophrenia+SA/D, Bipolar Disorder+SA/D,
Depression+SA/D, Perceives hidden threats in others
Contextual Factors: Victimized in past year, family or friend
died in past year, fired from job in past year, divorced or
separated in past year, Unemployed for past year
Top Ten Predictors of Any Violent
Behavior Between Waves 1 and2









Age (younger)
History of any violent act
Male
Divorce or separation in the past year
History of physical abuse
Parental criminal history
Unemployment for the past year
Co-occurring severe mental illness and substance
abuse
Victimization in the past year




SMI did not predict severe/serious violence, even
when combined with substance use disorders.
SMI was significantly associated with physical abuse
by parents, parental arrests, substance disorders,
recent victimization, and unemployment.
46% of those with SMI had co-morbid substance
abuse/dependence. Violence risk was higher in this
group than substance use without SMI.
People with SMI were more vulnerable to past
histories that elevate violence risk and more prone
to experience environmental stressors that also
elevate violence risk.



Severe mental illness did NOT rank among the
strongest predictors of violent behavior.
Severe mental illness alone was NOT statistically
related to future violence, in bivariate or
multivariate analyses.
People with any type of severe mental illness were
NOT at increased risk of committing serious/severe
violent acts.
Evolution from Violence Prediction
to Risk (or threat) Assessment


Violence Prediction
-focuses on the individual
-portrays dangerousness as a state
Risk Assessment
-focuses on person-situation interactions
-portrays dangerousness as
dynamic, contextual and continuous
We continue to be asked to
assess risk of violence:








Need for admission/ suitability for discharge
ER evaluations
Civil Commitment/ Release
Workplace/ school threats
Juvenile justice management
Sentencing/ Parole/ Probation/ Early
Release
Sex Offender Commitments
Specialty Court Treatment Plans
Approaches to Risk Assessment

Unstructured Clinical Judgment

Actuarial

Structured Professional Judgment

Anamestic
Clinical Judgment


More accurate than chance (Mossman
1994) AUC= .67
Does facilitate aspects of data
gathering and data interpretation
Actuarial




Formal / equation-formula-graph- table used
to arrive at a probability of some outcome
Objective, mechanistic, reproducible
combination of predictive factors, selected
and validated through empirical research
against known outcomes
BUT clinicians have not embraced this
Hard to go from the abstract to the individual
Structured Professional
Judgment
Presentation of specific risk factors
derived from broad review of literature
not specific data set- factors are well
operationalized so their applicability can
be coded: yes-possibly-no-/ multiple
data sources/ evaluator draws
conclusion weighing risk factors and
intensity of management
Anamestic
Process of gathering detailed
information about individual’s history of
violence. Question in detail about each
particular violent event (preceding,
subsequent, during)- thoughts, feelings
and behaviors - to identify risk and
protective factors that recur across
violent events- identify target
interventions
A Practical Guide to Risk
Assessment…



Build structure into your approach
Remember that violence is not a
common event, so prediction is not
easy
Start by thinking about the base rate
for your clinical situation
“Knowledge of the appropriate base rate
is the most important single piece of
information necessary to make an
accurate [violence] risk prediction.”
(Monahan 1981)
Base Rate



The proportion of a particular population
who commit violence in a particular period
of time
Starting point for subsequent evaluation of
probability
Varies by type of violence, by method of
detection, over time, and usually
underestimates the true extent of violence
Practical Assessment of potential
for violence toward others involves
considering…




Risk and Protective Factors: variables
associated with the probability that violence
will or will not occur
Harm: the nature and severity of the
probable results of the violent behavior
Risk Level: the probability that violence
will occur
Potential Victims: who are the likely
objects of the violence
Standardized Risk
Assessment Tools





Assist the Clinician in gathering appropriate data
Anchor assessment to established research
Access factors that are known to be associated
with particular types of violence in specific
populations.
Should be used in conjunction with clinical risk
assessment
May not be as objective in application as we
would hope.
When you think about using
Risk Assessment Tools:






Context
Purpose
Population
Parameters
Approach
Applicability
Heilbrun, et al. Violence Risk Assessment Tools: Overview and Analysis/
Otto and Douglas Handbook of Violence Risk Assessment
Structured Risk
Assessment Tools
Hare Psychopathy Checklists
(PCL, PCL-R, PCL:SV, PCL:YV)
 Historical-Clinical Mangement-20
(HCR-20) Violence Risk Assessment
Scheme
 COVR-Classification of Violence Risk

Hare Psychopathy Checklists
(PCL, PCL-R, PCL:SV, PCL:YV)






PCL-R: 20 item construct rating scale
Used in research and clinical settings
Assesses psychopathy in adults
Involves semi-structured interview and
review of file/collateral data
PCL-SV: 12 item
PCL:YV: 20 items
Psychopathic Personality
Construct
Personality traits and socially deviant behaviors:
 Glib and superficial charm
 Egocentricity
 Selfishness
 Lack of empathy, guilt and remorse
 Lack of enduring attachment to people, principles,
or goals
 Impulsive and irresponsible behavior
 Tendency to violate explicit social norms
PCL


Ability to predict violent behavior depends on type
of behavior being predicted( general v violent v
sexual), context in which offender is or will be
located ( corrections or community) and time frame
of prediction ( 1 or 10 years)- and demographic
variables – age/ gender/ race and ethnicity; need
specific referral question to determine if should be
used
Has modest to moderate relationship with future
community violence and weak to modest with
future institutional violence
Historical-Clinical Mangement-20 (HCR-20)
Violence Risk Assessment Scheme




Structured Professional Judgment model / translated into 16
languages
Intended to facilitate assessments of risk for interpersonal
violence (actual, attempted or threatened), clear unambiguous
threats of harm, including psychological harm, to person or
persons.
Intended to provide a structured assessment of the risk factors
that are present in a given case, the relevance of the risk factors
for a given individual’s violence risk, and what risk management
strategies might be put into place in order to mitigate that risk.
Historical / Clinical and Risk Management Scales
Checklist from HCR-20

Historical: previous violence, young
age at first violent incident,
relationship instability, employment
problems, substance use problems,
major mental illness, psychopathy,
early maladjustment, personality
disorder, prior supervision failure
Checklist from HCR-20


Clinical: lack of insight, negative
attitudes, active symptoms of major
mental illness, impulsivity,
unresponsiveness to treatment
Risk Management Items: plans lack
feasibility, exposure to destabilizers,
lack of personal support,
noncompliance with medication, stress
COVRClassification of Violence Risk




Computer-based program to determine
category of risk
Iterative Classification Tree (ICT)
Designed to mirror clinical decision
making process
Differentiates between low and high risk
populations
Violence Risk Assessment
Decision-Making Models
Flipping a Coin
Clinical Decision-Making
History of Violence
Psychopathy Checklist
Violence Risk Appraisal Guide
HCR-20
MacArthur Risk Assessment Study
->
->
->
->
->
->
->
AUC=.50
AUC=.66
AUC=.71
AUC=.75
AUC=.76
AUC=.80
AUC=.82
Perfect Accuracy
-> AUC=1.0
No risk assessment should rely solely on
the results of any one instrument:
information gleaned from these
structured instruments should be used
to inform risk assessment and assist in
risk communication.
Given all we now know…
What is a practical approach
to risk management?
LEAD: A Four-Step Approach to
Assessing Violence Risk




LOOK at static, individual-level factors known to
empirically relate to violent behavior in your
population (dispositional and historical factors)
EXAMINE for presence of protective factors or
unique individualized factors from both the micro
and macro environments (contextual factors)
ADJUST your risk assessment by considering
dynamic individual variables (clinical factors)
DOCUMENT your assessment and risk
management plan and communicate it to those
who need to know
Specific Tasks:
 Identifying/quantifying
the risk
 Modifying the acute risk
 Managing the chronic risk
 Balancing the seriousness of
potential outcome with the
probability of it’s occurrence
Keep in Mind:


Time frame of prediction
Structure of setting
– Institution v. Community

Impact of aging / group involvement
Think about identifying:



Behaviors that are not a product of illness
but likely to be patient choice
Lifestyle choices and issues that are going
to be difficult to modify and about which
you have no direct ability to modify
Patient’s competence to be making
decisions that might influence ability to
carry out violent act
Share the Risk – 4 C’s




Collect collateral information
Consultation with peers
Tackle limiting confidentiality head-on
Encourage cooperation of your patients and
their support systems in establishing risk
management plans and in managing risks
Develop a Violence
Prevention Plan





Distinguish Static ( Demographic and Past History)
and Dynamic Factors (Subject to change with
intervention such as access to weapons, psychotic
symptoms, active substance abuse, living setting
and situation)
Focus on current status of each dynamic factor
Develop a plan to address the combination of
factors unique to the individual
Determine the setting and parameters necessary to
safely implement the plan
Document this process
Communicate the risk and the
management plan to those who
need to know
Guided By:






Research Data / Instrument Construction
Ethical Standards
Laws
Admissibility Standards / Decisions
Professional Guidelines
Clinical experience
Manage your liability by the quality
of your risk assessment, the
thoughtfulness of your risk
management and the excellence
of your documentation.
Improving Your Understanding of
Violence and Mental Illness Just Might…




Help to eliminate destructive and common
myths about mental illness and violence
Reduce the overall incidence of violence in
our communities
Enhance safety in our clinical settings
Improve how the criminal justice system
responds to people with mental illness
Keep you out of the courtroom, where
the question for you could be…
Was your risk assessment
and your risk management…
Dead Right?
or
Dead Wrong?
The Jury Is Still Out!