PART I
Clinical Issues
1
Clinical Guidelines for the Assessment of Imminent Violence
ROBERT E. FEINSTEIN
Almost daily, psychiatric clinicians must evaluate the imminent risk of violence posed by their patients. Their decisions are frequently guided by nonspecific impressions or intuitive clinical judgments. Many psychiatric professionals feel discouraged when making assessments about violence because of an overly pessimistic reading of early studies on violence prediction. These studies (Cohen et al., 1978; Kozol & Boucher, 1972; Monahan, 1982) suggest that there are no valid or reliable predictors of long-term violence. Long-term violence is defined as violence occurring within periods of years. While it may be true that we will never be able to predict the specific timing or nature of the violence, it seems likely that we can greatly improve our ability to make reliable probability estimates of the risk of imminent violence. Imminent violence is defined as violence that occurs within a three-week period after an initial evaluation. This paper will attempt to outline specific clinical guidelines which may improve our ability to make short-term probability estimates of imminent violence. A brief review of some of the relevant literature will be presented from which specific clinical guidelines will be derived. Three short clinical cases will then be presented to illustrate the use of these guidelines.
REVIEW OF THE LITERATURE
The combined wisdom of authors such as Cohen et al. (1978), Kozol and Boucher (1972), Monahan (1981, 1982) and Steadman et al. (1978) suggests that we are poorly equipped to make reliable long-term predictions about violence. A simplified summary is that these studies showed that approximately two-thirds of predictions of future violence were inaccurate. In other words, the prisoners, violent sex offenders, and psychiatric patients whom they predicted would again become violent did not do so in two-thirds of the cases. These early studies, however, may not be directly applicable to the probability estimates of imminent violence. As Monahan states in his classic work (1981, 1982), these studies were examining long-term predictions of the behavior of institutionalized clients after discharge. These studies did not assess the reliability of short-term probability estimates of imminent violence in a psychiatric population.
Pokorny (1983) pointed out that suicide predictions may be unsuccessful because suicides are rare events and, as such, cannot be easily predicted without creating false positives. This same logic has been applied to homicide and violence prediction. However, Skodal and Karasu's paper (1978) and extrapolation from Silver and Yudofsky's Overt Aggression Scale (1987) seems to contradict this common viewpoint. If violence is defined broadly to include homicide, physical assault, destruction of property, and verbal violence, then violence is not a rare event at all. Violence, broadly defined, is quite common in the psychiatric and general population.
Despite the continuing controversies about our ability to predict violence over the long term, there are other authors who are more optimistic about making probability estimates of imminent violence. These authors (Warner, 1961; Werner, 1983; Baxter, 1968; Bengelsdorf, 1984) have moved away from trying to predict the occurrence of a specific violent episode at a specific time. Instead, they have tried to find variables that will aid in assessing the probability of violence, broadly defined, over a short time span of days or weeks. Warner (1961) showed significant evidence that the ability to care for oneself, the presence of family supports, danger potential, and treatment prognosis were reliable variables that correlate with the probability of imminent violence in psychiatric patients. Baxter et al. (1968) showed that the duration of the illness, previous psychiatric illness, the ability to communicate, and personal appearance may also be factors to help in our assessment of the short-term risk of violence. Monahan (1981, 1982) added past history of violence, family relations, and coexisting medical conditions to a growing list of relevant factors. Bengelsdorf et al. (1984) developed a “Crisis Triage Rating Scale” which was based on three factors: dangerousness (which was an assessment of suicide and violence potential); support systems; and the patient's ability to cooperate with treatment. These three factors were shown to be significant determinants of clinicians' decisions as to which patients needed to be committed to a hospital and which could be released and treated in the community. The work of Bleuler (1930), Fish (1967), and Yesavage (1983) suggests that a patient's ideation offers valuable clues as to the potential for imminent violence. Plutchik et al. (1985) proposed that poor impulse control, feelings of hostility, the triad of enuresis, firesetting and cruelty to animals in childhood, menstrual problems, young age, limited education, past history of violence, and repeated automobile infractions are variables which may help us in our short-term assessments of the potential for violence. Tinklenberg and Woodrow (1974) and Elliot (1987) added substance abuse to the list of relevant factors. Silver and Yudofsky (1987) remind us of the importance of neuropsychiatric disorders as risk factors for imminent violence. Finally, McNeil and Binder (1987) showed that emergency commitment did permit judgments of danger-ousness with a high degree of short-term predictive reliability.
CLINICAL GUIDELINES IN ASSESSING THE RISK OF IMMINENT VIOLENCE
Based on the above literature review, it is possible to propose the use of eight broad variables to assist clinicians in making probability estimates of imminent violence. These variables can be used as a guide when making clinical decisions about shortterm violence potential. These variables are: (1) violent ideation; (2) behavior during the interview; (3) recent history of violence; (4) past history of violence; (5) support network; (6) cooperation with treatment; (7) substance abuse; and (8) neurological or medical illnesses associated with violence. It is unlikely that any single variable, by itself, will be significantly useful in making probability estimates of imminent violence. The literature suggests that the complex interactions of the constellation of these eight variables may be the most useful approach to the clinical assessment of the potential for violence in the immediate future. Clinical application and assessment of imminent violence using these variables will therefore be described.
Violent Ideation
The literature suggests that patients' violent thought or homicidal ideation can probably be put in a spectrum, from a higher to a lower probability of violence risk. The highest-risk thoughts on this continuum suggesting a high probability of imminent violence are those involving an intense desire to kill a specific person. Other work (Goodwin et al., 1971; Hellerstein et al., 1987) suggests that command hallucinations and/or delusions to hurt or kill are also the highest-risk ideation for imminent violence. A nonspecific wish to hurt or kill a member of a group (i.e., a wish to kill blacks or homosexuals) probably carries a slightly more modest risk for short-term violence potential. Ambivalent wishes to hurt others or damage objects, or nonspecific feelings of hostility, may carry lesser risks. Patients who have few thoughts of violence or homicide have the least risk of short-term violence potential.
Behavior During the Interview
My earlier work (Feinstein, 1986) suggests that there is a natural progression of behaviors that signal increasing risk of violence during the interview. Patients progress from a calm phase to an increase in psychomotor movements followed by an “early verbal phase” where they question authority. Increasing risk is signaled by a “late verbal phase” where patients may challenge authority, use profanity, or display approach-avoidance behaviors which frequently lead to a violent assault. By observing these behaviors as patients progress through them, it is possible to anticipate the immediate risk of violence and take preventative measures.
Recent History of Violence
There may be a similar spectrum of recent events or recent behaviors which may correlate with the degree of imminent violence potential. Violence risk, from the highest to the lowest probability for potential violence, includes: (a) impulsive or deliberate assaultive behavior with weapons (e.g., assaults with a gun, knife, club, or bottle); (b) impulsive or deliberate physical assaults (e.g., beatings or an attempt at strangulation which have caused either fractures or the need for hospital assessment); (c) recent violence without serious sequellae (e.g., a slap, punch, or push); (d) other impulsive or unpredictable behaviors indicating a moderate risk potential for violence; (e) destruction of property or isolated objects, which may indicate a lower risk potential for violence; (f) no recent behavior associated with violence and clinical evidence that a patient's impulses are well regulated, indicating only minimal risk for imminent violence.
Past History of Violence
The past history of violence is still considered by many to be the best indicator of the probability of short-term risk for violence. A detailed history about a patient's past pattern of violence is essential for the assessment of imminent violence potential, as well as for planning interventions which may help prevent violence. Special attention should be paid to previous precipitants that initiated violence, the severity of past violence, the frequency of the violence, and the countervailing forces (Plutchik et al., 1985) which may inhibit or attenuate violence. Psychiatric patients who repeatedly become violent generally show repetitive patterns to their violence. As part of this assessment, clinicians should inquire about a patient's knowledge of weapons and their availability.
It is also useful to obtain a history of past arrests, automobile infractions, criminal records, or involvement in frequent legal proceedings. A childhood history of frequent disruptive changes in caretakers (Rockwell, 1972), frequent childhood abuse or punishment (Bryer et al., 1987), and the triad of enuresis, firesetting and cruelty to animals (Plutchik et al., 1985; Felthous et al., 1987) may also be associated with an increased potential for imminent violence. Also, adolescent histories of chronic problems with authority (as evidenced by truancy, running away, losing one's temper easily, sexual promiscuity, and overeating) (Plutchik et al., 1985) may also carry a greater risk in the short term for future violence.
There is little convincing evidence as to the relative significance of such past history regarding the probability of imminent violence potential.
Support Systems
Bengelsdorf et al. (1984), Schoenfeld (1986), Schnur et al. (1986) and new research (Feinstein et al., 1988) indicate that the social network or support system may be the main variable used by emergency psychiatric clinicians when determining which violent patients need commitment and which ones can be released. Support systems such as family, friends, mental health care providers, religious groups, etc. should be evaluated according to their interest, availability and competence. There is probably a spectrum of quality among support networks. A competent support system that is disinterested or unavailable will increase the likelihood of imminent violence. Similarly, an interested and available support system which is not competent (i.e. children) increases the likelihood for short-term violence potential. Patients who are discharged to support systems that demonstrate competence, interest and availability are at only slight risk for imminent violence.
Ability to Cooperate with Treatment
This variable is widely used by clinicians in deciding if a potentially violent patient can be released. The research of Feinstein, Plutchik, and van Praag (1988) on 95 patients evaluated at the Bronx Municipal Hospital /Albert Einstein College of Medicine confirms the earlier work of Bengelsdorf (1984). It demonstrated a high correlation between a patient's ability to cooperate with treatment and clinicians' judgments about admitting a potentially violent patient. Patients who refused to cooperate with their treatment or who were unable to do so in the past were likely to be admitted. Patients who had weak motivation for treatment or who showed limited capacity to participate were less frequently admitted. Patients who actually sought treatment or who demonstrated a willingness or ability to participate in their treatment were more typically treated in outpatient settings.
Substance Abuse History
The literature correlating acute alcohol intoxication and an acute risk of imminent violence is extensive. The correlation between other fo...