- Email: [email protected]
Violence and suicide risk assessment in the psychiatric emergency room
Violence and Suicide Risk Assessment Psychiatric Emergency Room
in the
Robert Feinstein and Robert F’lutchik Structured clinical rating scales covering 10 areas related to suicide and violence were constructed for use in a psychiatric emergency room (ER). Ninety-five ER patients were evaluated with the scales, 50 of whom were discharged after the visit and 45 of whom were admitted to the inpatient psychiatric wards of the hospital. The admitted patients were found to differ significantly from the discharged patients on every one of the 10 scales. Scores on the scales were also found to predict suicide precautions on the wards, harrassment of other patients as assessed from nursing notes, and indicators of violence on the wards. The scales were also found to have high internal reliability and high sensitivity and specificity. They appear to be helpful to clinicians in identifying patients in need of hospitalization and may also serve as limited predictors of hospital functioning. 0 1990 by W. B. Saunders Company.
A
LMOST ALL PATIENTS admitted to inpatient psychiatric wards are screened or evaluated in a psychiatric emergency room, general emergency room (ER), or by other admitting services or clinicians. In most such settings, only a small number of symptoms determine whether a patient is admitted or discharged. Such factors as the patient’s ability to care for self, presence of family supports, danger potential and treatment prognosis have been used,’ as well as duration of illness, previous illnesses, ability to communicate, and personal appearance.2 Bengelsdorf et a1.3 developed a crisis triage rating system that is based on three factors: dangerousness, support system, and motivation or ability to cooperate. These investigators reported that scores based on such ratings were 97% concordant with decisions made by a crisis team on the basis of clinical judgement. A comparison of voluntary and two-physician (2-PC) committed patients at a state hospital showed that only a small number of symptoms distinguished between the two groups. These symptoms were antisocial acts, anger, belligerence, negativism, agitation, and assaultive acts.4 Evidently, these symptoms are largely reflections of dysfunctions of aggression. In psychiatric ERs, violent ideation or acts are not uncommon. Based on a random sample of 367 psychiatric ER patients, Skodal and Karasu’ found that 17% of the patients were described as violent because of outwardly directed aggressive ideation or behavior in their clinical presentations, and another 17% had suicidal tendencies without aggression directed toward others. Approximately 5% of the sample was both violent toward others and suicidal. It is of interest to note that of the cases defined as violent in the ER, 70% had acted out violent impulses before their ER visit. Of these “repeaters” approximately 70% denied any degree of premeditation before acting, thus implying an important role of impulsivity in violence. An earlier report on self-directed violence in ER patients indicated that 28% were suicidal.6
From the Department of Psychiatry, Albert Einstein College of Medicine/Monte$ore Bronx, NY. Address reprint requests to Robert Feinstein. M.D., 212 New Canaan Ave. Norwalk, G 1990 by W. B. Saunders Company. 00~0-440x~90/3104-~005$03.00~0
Comprehensive
Psychiatry,
Vol. 3 1, NO. 4 (July/August),
1990: pp 337-343
Medical Center CT 06850.
337
338
FEINSTEIN AND PLUTCHIK
With the general increase in crimes of violence over the past several decades, and with complex legal issues raised by the Tarasoff case’ and others,8 there has been increased concern with identifying violent individuals in the ER.9 Also of great interest has been the question of how well psychiatrists can predict violent behaviors,” although most commentators on this issue have concluded that prediction is poor. The problem is complex and far from being resolved. The present study is concerned with three questions. First, what are the differences on measures of violence and other indices, between ER patients who are admitted to the inpatient service and those who are discharged? Second, to what extent do measures of violence obtained in the ER correlate with indices of violence for these same patients on the inpatient wards? Third, to what extent do measures of suicidal behaviors obtained in the ER correlate with indices of suicidal behavior for the same patient on the inpatient wards? METHODS Based on an extensive review of the literature, a Violence and Suicide Assessment form (VASA) was constructed for the study. This form, shown in Table 1, covers 10 areas of interest. These are current violent thoughts (during interview), recent violent behaviors (during the past several weeks), past history of violent/antisocial/disruptive behaviors (lifetime history), current suicidal thoughts, recent suicidal behaviors, past history of suicidal behaviors, support systems, ability to cooperate, substance abuse, and reactions during the interview. Within each area of interest there are a number of brief descriptions of relevant behaviors varying in degree of severity or degree of psychopathology. For example, under the area of current violent thoughts, the items and weights given to them are as follows: 4, expresses intense wish to kill someone specific; 3, reveals command hallucinations to injure someone; 2, expresses ambivalent wish to kill someOne specific; 1, expresses nonspecific feelings of rage or belligerence; and 0, reveals no homicidal ideas. At the end of the rating scales, the clinician is asked to make a probability estimate of the likelihood of suicidal ideation or behavior and a separate probability estimate of the likelihood of violent ideation or behavior. The instructions given the clinicians are: “Your probability estimate (on a scale from 0 to 100) should refer to the next 3 weeks. In that period, do you expect that this patient will show suicidal ideation or behavior and/or violent ideation or behavior?” This VASA form was used in the ER of a large municipal hospital by clinicians during a 4-month period. During this time they evaluated 95 patients on the form as part of their usual screening, evaluation, and treatment functions. The evaluations and decisions on a patient were made first. Then the VASA form was completed. Patients in the study were selected from the psychiatric emergency service, subject only to the need for voluntary cOnsent and being over the age of 1S years. The design of the study was based on the idea that some of the ER patients would be hospitalized and most would be discharged home or referred for further outpatient care. The final sample consisted of SO discharged patients and 45 patients admitted to the inpatient wards. Since the average length of stay of psychiatric patients at this hospital is approximately 3 weeks, all of the admitted patients were discharged by the time the follow-up data collection period was instituted. For each hospitalized patient, information was obtained on his (or her) number of seclusions, reasons for seclusion, the number and nature of incident reports, the presence of suicide precautions, diagnoses, discharge disposition, drugs used, and nursing notes relevant to suicidal or violent behavior. Based on this information. several indices were constructed of violence-related or suicide-related ward behavior.
RESULTS
Table 2 presents the general demographic data for the patients, including sex distribution, mean age, marital status, race, and diagnoses. The two groups have approximately the same age and racial distribution, but the inpatient group has relatively more single males. Inpatients also have a higher frequency of multiple suicide attempts, more of a history of drug abuse, and more signs of suicidality (gestures and ideation) at the time of admission. Twenty-three
Table 1. Violence
and Suicide Assessment
Scale
Current Violent Thoughts (during interview) Expresses intense wish to kill someone specific. 4 Reveals command hallucinations to injure someone. 3 Expresses ambivalent wish to kill someone specific. 2 1 Expresses nonspecific feelings of rage and belligerence. Reveals no homicidal ideas. 0 Recent Violent Behaviors (during the past several weeks) Showed serious assaultive behavior (e.g., tried to strangle, stab, or shoot someone). 4 Beat up someone badly (e.g., broke bones or required hospitalization). 3 2 Slapped or pushed or punched someone (no serious outcomes). 1 Broke things in house or elsewhere. Showed good control of his (her) behavior. 0 Past History of Violent/Antisocial/Disruptive Behaviors (lifetime history) 4 Has committed violent acts in the past (e.g., beaten up people). 4 Has been arrested for assaultive behavior. 3 Carries weapons (e.g., knife, gun, chain, razor, etc.). Has access to weapons. 3 Has been arrested for automobile infractions. 2 Has a criminal record. 2 2 Chronic problems with authority (e.g., truancy, running away from home, family fights). 2 Has a history of impulsive or unpredictable behavior. (e.g., loses temper easily, overeats, sexual promiscuity, etc.). Frequent changes of living situation as a child. 2 Has no past history of violence. 0 Current Suicidal Thoughts (during interview) Expresses intense wish to kill self and has made a plan. 4 Reveals psychotic or delusional ideation or hallucinations to kill or injure self. 4 3 Expresses intense wish to kill self but has made no plan. 2 Expresses ambivalent wish to kill self. 0 Reveals no suicidal ideas. Recent Suicidal Behaviors (during the past several weeks) 4 Made a serious suicidal attempt (e.g., tried to kill self by gunshot, ingestion, hanging or jumping). 3 Made a suicidal gesture (e.g., superficially cut wrist or ingested two pills). 3 Made a specific suicide plan. 3 Attempt made with little chance of discovery by others. 2 Had no interest or hope for the future. 0 Has no suicidal plans or attempts. Past History of Suicide (lifetime history) 4 Mother, father or sibling has committed suicide or made a suicide attempt. 3 Has (or had) a diagnosis of major affective disorder or psychosis. 3 Has made one or more previous suicide attempts. 2 Current attempt is an “anniversary” reaction. 2 Has a serious medical illness or disability. 0 Has no past history of suicidal ideas or attempts. Support Systems/Stresses 3 No family, friends, social agency, or psychiatrist available. 2 Has tenuous connection with family, friends, social agency, or psychiatrist. 2 Has had many recent life stresses (e.g., job, family, children, health, etc.) 1 Has a family which is marginally willing or able to help. 0 Has a family strongly committed and able to help. Ability to Cooperate 3 Refuses to cooperate with interview and treatment plan. 2 Unable to cooperate with interview and treatment plan. 1 Wants help but motivation is weak. 0 Actively seeks treatment: willing and able to cooperate. Substance Abuse 3 Is intoxicated. 3 Is in withdrawal. 3 Is a compulsive long-term drug abuser (includes alcohol or other drugs). 2 Is an occasional drug abuser (alcohol or other drugs). 1 Recreational use of drugs. 0 No abuse of any drugs. Reactions During Interview 4 Assaultive behavior against a person (or object) in the environment. 3 Challenges authority (e.g., curses, yells, screams). 2 Shows approach-avoidance behavior toward interviewer. 1 Shows motoric activity (e.g.. pacing, smoking, fidgeting, etc.). 1 Seems very impatient. Calm, seated, responsive to questions. 0 m 1986 by Robert Feinstein and Robert Plutchik.
FEINSTEIN AND PLUTCHIK
340
Table 2. Demographic Information Comparing the ER Patients Who Were Inpatients With Those Who Were Discharged from the ER
Admitted
Inpatients (N = 45)
Outpatients (N = 50)
Mean age (yr)
33.4
34.2
Male Female
62% 38%
38% 62%
Single Married Divorced Separated Widowed
66% 14% 11% 9% 0%
51% 26% 9% 5% 9%
35% 23% 40% 2%
38% 28% 34% 0%
5% 9% 86%
0% 10% 90%
38% 7% 14% 41%
27% 2% 6% 27%
Brought by police
23%
8%
Major depression Dysthymia Schizophrenia spectrum Substance abuse
25% 3% 42% 31%
20% 15% 30% 35%
Variables
White Black Hispanic Other More than one prior suicide attempt One prior suicide attempt No prior suicide attempt History Suicide Suicide Suicide
of substance abuse (yes) attempt at admission (yes) gesture at admission (yes) ideation at admission (yes)
as
percent of the inpatients were brought to the ER by the police, while only 8% of
VIOLENCE/SUICIDE
RISK ASSESSMENT
IN THE ER
341
Table 3 presents a comparison of ER patients who were discharged with those who were admitted to the inpatient wards. The admitted patients are significantly more functionally impaired than the discharged patients on every variable. They are more violent, more suicidal, less cooperative, more substance abusing, and with fewer social supports. The clinicians rate them as having a significantly higher probability of both suicide and violence, and they have a significantly greater history of prior suicide attempts and violent episodes. Following is a summary of the significant product-moment correlations obtained in the study that describe short-term predictions. The presence of suicidal ideation at the time of admission to the ER is highly correlated with suicide precautions taken on the ward (+.63) and with the number of days on suicide precautions (+ .32). The number of prior suicide attempts is highly correlated (+ .61) with the harrassment of others on the wards (as judged from nursing notes), and is also highly correlated with the risk of violence in the hospital (+ .60). The risk of violence in the hospital as a composite score based on the sum of the following 10 items: total number of seclusions, uncontrolled behavior, physical aggression against others, loud noisy behavior, inappropriate social behavior, destructive toward objects, verbal threats or abuse, violence toward staff, violence toward patients, and harrassment of others. Information about these items was obtained directly from the patients’ medical records. The risk of violence in the hospital is also highly correlated with a history of substance abuse (+.63). Finally, it appears that the suicide probability estimate is correlated with the violence probability estimate ( + S3). Since the lo-item VASA scale appears to have high internal reliability, its sensitivity and specificity were also determined. The ability of the total scale score to Table 3. Comparison
of ER Patients With Those Admitted to the Inpatient the Items of the VASA Scale
Service Using
Discharged Patients (N = 50)
Admitted Patients (N = 45)
Items
Mean
SD
Mean
SD
t
Current violent thoughts Recent violent behavior Past history of violence Current suicidal thoughts Recent suicidal behaviors Past history of suicide Support systems/stresses Ability to cooperate Substance abuse Negative reactions during interview
.12 .38 1.36 .40 .44 .a8 1 .oo .4a .a2
.48 .92 2.83 .a1 1.03 1.61 1.18 .71 1.49
.a4 1.09 2.91 2.07 2.13 3.35 2.27 2.13 1.73
i .38 1.36 3.09 2.77 1.89 2.59 1.56 4.38 I .a2
3.34t 2.92t 2.52t 4.81$ 5.33% 5.40$ 4.433 2.50* 2.62*
.36
.92
1.84
2.50
3.75$
Total score Suicide probability estimate Violence probability estimate
6.24 20% 12%
6.71 17.3 14.8
19.38 52% 44%
9.26 30.5 30.6
7.84$ 6.15$ 6.154
lP < .05. tP< .Ol. SP< .OOl.
FEINSTEIN AND PLUTCHIK
342
discriminate between those ER patients who were admitted to the inpatient service and those who were not was examined for different cutoff scores. As typically found, when specificity (i.e., the ability to identify true negatives) increased, sensitivity (e.g., the ability to identify true positives) decreased. The optimum cutoff score for maximum sensitivity and specificity was found to be 11. A score of this value produced a sensitivity and specificity of approximately 82%. DISCUSSION
The focus of this study has been on the initial assessment of violence or suicide risk in a psychiatric ER and the extent to which such assessments are able to predict certain aspects of the subsequent course of hospitalization. It was found that patients who are immediately hospitalized after an initial ER evaluation are somewhat different from those who are discharged home. They tend to be younger, contain a higher proportion of males, are more likely to be single, have a more frequent history of substance abuse, are more likely to have made multiple suicide attempts in the past, and are more likely to present with suicidal ideation or gestures. All patients were assessed using an expanded lo-item version of the Bengelsdorf et al (1984) triage rating system. The results showed that the ER patients admitted to the inpatient wards scored significantly higher on every one of the items as well as on total score. They also scored higher on the frequency of both prior suicide attempts and violent episodes. These findings clearly indicate that the VASA scale significantly discriminates between admitted and discharged ER patients. As further support of this point, the sensitivity and specificity were both found to be quite high (82%) using a cutoff score of 11. It thus appears that the VASA scale may be a useful tool for clinicians in ERs to guide the clinical interview. The second key question of the study was to examine whether initial ER data have any predictive value in describing aspects of the course of short-term inpatient hospitalization. The results indicate that the number of prior suicide attempts is highly correlated with the likelihood of violent behavior in the hospital (r = .60). A history of substance abuse is also highly correlated with the likelihood of violent behavior in the hospital within 3 or 4 weeks after admission (r = .63). It was also found that three items of the VASA scale (lifetime history of suicide attempts, lack of social support systems, and inability to cooperate with the interviewer) correlate significantly with suicide risk in the hospital (r = .41) as estimated from nursing notes. It thus appears that some degree of prediction of acting out aggressive behavior is possible over relatively short time periods, using simple clinically obtained measures. We thus believe that the rating scales described here have value for the clinician to help in the identification of patients who need to be hospitalized, and may also serve as limited predictors of some aspects of hospital functioning. ACKNOWLEDGMENT The authors would like to thank Janie Lynn Feldman for assistance in data collection.
REFERENCES 1. Warner SL: Criteria for involuntary hospitalization of psychiatric patients in a public hospital. Ment Hygiene45:122-128,196l
VIOLENCE/SUICIDE
RISK ASSESSMENT
IN THE ER
343
2. Baxter S, Chodoroff B, Underhill R: Psychiatric emergencies dispositional determinants and the validity of the decision to admit. Am J Psychiatry 124:1542-1548, 1968 3. Bengelsdorf H, Levy LE, Emerson RL, et al: A crisis triage rating scale: Brief dispositional assessment of patients at risk for hospitalization. J Nerv Ment Dis 172:424-430, 1984 4. Zwerling I, Karasu TB, Plutchik R, et al: A comparison of voluntary and involuntary patients in a state hospital. Am J Orthopsychiatry 45:81-86, 1975 5. Skodal CH, Karasu TB: Emergency psychiatry and the assaltive patient. Am J Psychiatry 135:202-205, 1978 6. Browning CH, Tyson RL, Miller SI: A study of psychiatric emergencies. Part II. Suicide Psychiatry Med 2:359-366, 1971 7. Tarasoff V: Regents of the University of California et al: 13 1 Cal Rptr 14, 17 Cal 3d 425, 55 1 Pzd 334 (1976) 8. Rachlin S, Schwartz HI: Unforseeable liability for patients’ violent acts. Compr Psychiatry 371731-775, 1986 9. Feinstein R: Clinical guidelines for the assessment of immanent violence, in van Praag H, Plutchik R, Apter A (eds): Violence and Suicidality: Perspectives in Clinical and Psychobiological Research. New York, NY, Bruner Maze], 1990 10. Monahan J: The clinical prediction of violent behavior. US Department of Health and Human Services Publication No. ADM 8 l-921. Washington, DC, US Government Printing Office, 198 1
in the
Robert Feinstein and Robert F’lutchik Structured clinical rating scales covering 10 areas related to suicide and violence were constructed for use in a psychiatric emergency room (ER). Ninety-five ER patients were evaluated with the scales, 50 of whom were discharged after the visit and 45 of whom were admitted to the inpatient psychiatric wards of the hospital. The admitted patients were found to differ significantly from the discharged patients on every one of the 10 scales. Scores on the scales were also found to predict suicide precautions on the wards, harrassment of other patients as assessed from nursing notes, and indicators of violence on the wards. The scales were also found to have high internal reliability and high sensitivity and specificity. They appear to be helpful to clinicians in identifying patients in need of hospitalization and may also serve as limited predictors of hospital functioning. 0 1990 by W. B. Saunders Company.
A
LMOST ALL PATIENTS admitted to inpatient psychiatric wards are screened or evaluated in a psychiatric emergency room, general emergency room (ER), or by other admitting services or clinicians. In most such settings, only a small number of symptoms determine whether a patient is admitted or discharged. Such factors as the patient’s ability to care for self, presence of family supports, danger potential and treatment prognosis have been used,’ as well as duration of illness, previous illnesses, ability to communicate, and personal appearance.2 Bengelsdorf et a1.3 developed a crisis triage rating system that is based on three factors: dangerousness, support system, and motivation or ability to cooperate. These investigators reported that scores based on such ratings were 97% concordant with decisions made by a crisis team on the basis of clinical judgement. A comparison of voluntary and two-physician (2-PC) committed patients at a state hospital showed that only a small number of symptoms distinguished between the two groups. These symptoms were antisocial acts, anger, belligerence, negativism, agitation, and assaultive acts.4 Evidently, these symptoms are largely reflections of dysfunctions of aggression. In psychiatric ERs, violent ideation or acts are not uncommon. Based on a random sample of 367 psychiatric ER patients, Skodal and Karasu’ found that 17% of the patients were described as violent because of outwardly directed aggressive ideation or behavior in their clinical presentations, and another 17% had suicidal tendencies without aggression directed toward others. Approximately 5% of the sample was both violent toward others and suicidal. It is of interest to note that of the cases defined as violent in the ER, 70% had acted out violent impulses before their ER visit. Of these “repeaters” approximately 70% denied any degree of premeditation before acting, thus implying an important role of impulsivity in violence. An earlier report on self-directed violence in ER patients indicated that 28% were suicidal.6
From the Department of Psychiatry, Albert Einstein College of Medicine/Monte$ore Bronx, NY. Address reprint requests to Robert Feinstein. M.D., 212 New Canaan Ave. Norwalk, G 1990 by W. B. Saunders Company. 00~0-440x~90/3104-~005$03.00~0
Comprehensive
Psychiatry,
Vol. 3 1, NO. 4 (July/August),
1990: pp 337-343
Medical Center CT 06850.
337
338
FEINSTEIN AND PLUTCHIK
With the general increase in crimes of violence over the past several decades, and with complex legal issues raised by the Tarasoff case’ and others,8 there has been increased concern with identifying violent individuals in the ER.9 Also of great interest has been the question of how well psychiatrists can predict violent behaviors,” although most commentators on this issue have concluded that prediction is poor. The problem is complex and far from being resolved. The present study is concerned with three questions. First, what are the differences on measures of violence and other indices, between ER patients who are admitted to the inpatient service and those who are discharged? Second, to what extent do measures of violence obtained in the ER correlate with indices of violence for these same patients on the inpatient wards? Third, to what extent do measures of suicidal behaviors obtained in the ER correlate with indices of suicidal behavior for the same patient on the inpatient wards? METHODS Based on an extensive review of the literature, a Violence and Suicide Assessment form (VASA) was constructed for the study. This form, shown in Table 1, covers 10 areas of interest. These are current violent thoughts (during interview), recent violent behaviors (during the past several weeks), past history of violent/antisocial/disruptive behaviors (lifetime history), current suicidal thoughts, recent suicidal behaviors, past history of suicidal behaviors, support systems, ability to cooperate, substance abuse, and reactions during the interview. Within each area of interest there are a number of brief descriptions of relevant behaviors varying in degree of severity or degree of psychopathology. For example, under the area of current violent thoughts, the items and weights given to them are as follows: 4, expresses intense wish to kill someone specific; 3, reveals command hallucinations to injure someone; 2, expresses ambivalent wish to kill someOne specific; 1, expresses nonspecific feelings of rage or belligerence; and 0, reveals no homicidal ideas. At the end of the rating scales, the clinician is asked to make a probability estimate of the likelihood of suicidal ideation or behavior and a separate probability estimate of the likelihood of violent ideation or behavior. The instructions given the clinicians are: “Your probability estimate (on a scale from 0 to 100) should refer to the next 3 weeks. In that period, do you expect that this patient will show suicidal ideation or behavior and/or violent ideation or behavior?” This VASA form was used in the ER of a large municipal hospital by clinicians during a 4-month period. During this time they evaluated 95 patients on the form as part of their usual screening, evaluation, and treatment functions. The evaluations and decisions on a patient were made first. Then the VASA form was completed. Patients in the study were selected from the psychiatric emergency service, subject only to the need for voluntary cOnsent and being over the age of 1S years. The design of the study was based on the idea that some of the ER patients would be hospitalized and most would be discharged home or referred for further outpatient care. The final sample consisted of SO discharged patients and 45 patients admitted to the inpatient wards. Since the average length of stay of psychiatric patients at this hospital is approximately 3 weeks, all of the admitted patients were discharged by the time the follow-up data collection period was instituted. For each hospitalized patient, information was obtained on his (or her) number of seclusions, reasons for seclusion, the number and nature of incident reports, the presence of suicide precautions, diagnoses, discharge disposition, drugs used, and nursing notes relevant to suicidal or violent behavior. Based on this information. several indices were constructed of violence-related or suicide-related ward behavior.
RESULTS
Table 2 presents the general demographic data for the patients, including sex distribution, mean age, marital status, race, and diagnoses. The two groups have approximately the same age and racial distribution, but the inpatient group has relatively more single males. Inpatients also have a higher frequency of multiple suicide attempts, more of a history of drug abuse, and more signs of suicidality (gestures and ideation) at the time of admission. Twenty-three
Table 1. Violence
and Suicide Assessment
Scale
Current Violent Thoughts (during interview) Expresses intense wish to kill someone specific. 4 Reveals command hallucinations to injure someone. 3 Expresses ambivalent wish to kill someone specific. 2 1 Expresses nonspecific feelings of rage and belligerence. Reveals no homicidal ideas. 0 Recent Violent Behaviors (during the past several weeks) Showed serious assaultive behavior (e.g., tried to strangle, stab, or shoot someone). 4 Beat up someone badly (e.g., broke bones or required hospitalization). 3 2 Slapped or pushed or punched someone (no serious outcomes). 1 Broke things in house or elsewhere. Showed good control of his (her) behavior. 0 Past History of Violent/Antisocial/Disruptive Behaviors (lifetime history) 4 Has committed violent acts in the past (e.g., beaten up people). 4 Has been arrested for assaultive behavior. 3 Carries weapons (e.g., knife, gun, chain, razor, etc.). Has access to weapons. 3 Has been arrested for automobile infractions. 2 Has a criminal record. 2 2 Chronic problems with authority (e.g., truancy, running away from home, family fights). 2 Has a history of impulsive or unpredictable behavior. (e.g., loses temper easily, overeats, sexual promiscuity, etc.). Frequent changes of living situation as a child. 2 Has no past history of violence. 0 Current Suicidal Thoughts (during interview) Expresses intense wish to kill self and has made a plan. 4 Reveals psychotic or delusional ideation or hallucinations to kill or injure self. 4 3 Expresses intense wish to kill self but has made no plan. 2 Expresses ambivalent wish to kill self. 0 Reveals no suicidal ideas. Recent Suicidal Behaviors (during the past several weeks) 4 Made a serious suicidal attempt (e.g., tried to kill self by gunshot, ingestion, hanging or jumping). 3 Made a suicidal gesture (e.g., superficially cut wrist or ingested two pills). 3 Made a specific suicide plan. 3 Attempt made with little chance of discovery by others. 2 Had no interest or hope for the future. 0 Has no suicidal plans or attempts. Past History of Suicide (lifetime history) 4 Mother, father or sibling has committed suicide or made a suicide attempt. 3 Has (or had) a diagnosis of major affective disorder or psychosis. 3 Has made one or more previous suicide attempts. 2 Current attempt is an “anniversary” reaction. 2 Has a serious medical illness or disability. 0 Has no past history of suicidal ideas or attempts. Support Systems/Stresses 3 No family, friends, social agency, or psychiatrist available. 2 Has tenuous connection with family, friends, social agency, or psychiatrist. 2 Has had many recent life stresses (e.g., job, family, children, health, etc.) 1 Has a family which is marginally willing or able to help. 0 Has a family strongly committed and able to help. Ability to Cooperate 3 Refuses to cooperate with interview and treatment plan. 2 Unable to cooperate with interview and treatment plan. 1 Wants help but motivation is weak. 0 Actively seeks treatment: willing and able to cooperate. Substance Abuse 3 Is intoxicated. 3 Is in withdrawal. 3 Is a compulsive long-term drug abuser (includes alcohol or other drugs). 2 Is an occasional drug abuser (alcohol or other drugs). 1 Recreational use of drugs. 0 No abuse of any drugs. Reactions During Interview 4 Assaultive behavior against a person (or object) in the environment. 3 Challenges authority (e.g., curses, yells, screams). 2 Shows approach-avoidance behavior toward interviewer. 1 Shows motoric activity (e.g.. pacing, smoking, fidgeting, etc.). 1 Seems very impatient. Calm, seated, responsive to questions. 0 m 1986 by Robert Feinstein and Robert Plutchik.
FEINSTEIN AND PLUTCHIK
340
Table 2. Demographic Information Comparing the ER Patients Who Were Inpatients With Those Who Were Discharged from the ER
Admitted
Inpatients (N = 45)
Outpatients (N = 50)
Mean age (yr)
33.4
34.2
Male Female
62% 38%
38% 62%
Single Married Divorced Separated Widowed
66% 14% 11% 9% 0%
51% 26% 9% 5% 9%
35% 23% 40% 2%
38% 28% 34% 0%
5% 9% 86%
0% 10% 90%
38% 7% 14% 41%
27% 2% 6% 27%
Brought by police
23%
8%
Major depression Dysthymia Schizophrenia spectrum Substance abuse
25% 3% 42% 31%
20% 15% 30% 35%
Variables
White Black Hispanic Other More than one prior suicide attempt One prior suicide attempt No prior suicide attempt History Suicide Suicide Suicide
of substance abuse (yes) attempt at admission (yes) gesture at admission (yes) ideation at admission (yes)
as
percent of the inpatients were brought to the ER by the police, while only 8% of
VIOLENCE/SUICIDE
RISK ASSESSMENT
IN THE ER
341
Table 3 presents a comparison of ER patients who were discharged with those who were admitted to the inpatient wards. The admitted patients are significantly more functionally impaired than the discharged patients on every variable. They are more violent, more suicidal, less cooperative, more substance abusing, and with fewer social supports. The clinicians rate them as having a significantly higher probability of both suicide and violence, and they have a significantly greater history of prior suicide attempts and violent episodes. Following is a summary of the significant product-moment correlations obtained in the study that describe short-term predictions. The presence of suicidal ideation at the time of admission to the ER is highly correlated with suicide precautions taken on the ward (+.63) and with the number of days on suicide precautions (+ .32). The number of prior suicide attempts is highly correlated (+ .61) with the harrassment of others on the wards (as judged from nursing notes), and is also highly correlated with the risk of violence in the hospital (+ .60). The risk of violence in the hospital as a composite score based on the sum of the following 10 items: total number of seclusions, uncontrolled behavior, physical aggression against others, loud noisy behavior, inappropriate social behavior, destructive toward objects, verbal threats or abuse, violence toward staff, violence toward patients, and harrassment of others. Information about these items was obtained directly from the patients’ medical records. The risk of violence in the hospital is also highly correlated with a history of substance abuse (+.63). Finally, it appears that the suicide probability estimate is correlated with the violence probability estimate ( + S3). Since the lo-item VASA scale appears to have high internal reliability, its sensitivity and specificity were also determined. The ability of the total scale score to Table 3. Comparison
of ER Patients With Those Admitted to the Inpatient the Items of the VASA Scale
Service Using
Discharged Patients (N = 50)
Admitted Patients (N = 45)
Items
Mean
SD
Mean
SD
t
Current violent thoughts Recent violent behavior Past history of violence Current suicidal thoughts Recent suicidal behaviors Past history of suicide Support systems/stresses Ability to cooperate Substance abuse Negative reactions during interview
.12 .38 1.36 .40 .44 .a8 1 .oo .4a .a2
.48 .92 2.83 .a1 1.03 1.61 1.18 .71 1.49
.a4 1.09 2.91 2.07 2.13 3.35 2.27 2.13 1.73
i .38 1.36 3.09 2.77 1.89 2.59 1.56 4.38 I .a2
3.34t 2.92t 2.52t 4.81$ 5.33% 5.40$ 4.433 2.50* 2.62*
.36
.92
1.84
2.50
3.75$
Total score Suicide probability estimate Violence probability estimate
6.24 20% 12%
6.71 17.3 14.8
19.38 52% 44%
9.26 30.5 30.6
7.84$ 6.15$ 6.154
lP < .05. tP< .Ol. SP< .OOl.
FEINSTEIN AND PLUTCHIK
342
discriminate between those ER patients who were admitted to the inpatient service and those who were not was examined for different cutoff scores. As typically found, when specificity (i.e., the ability to identify true negatives) increased, sensitivity (e.g., the ability to identify true positives) decreased. The optimum cutoff score for maximum sensitivity and specificity was found to be 11. A score of this value produced a sensitivity and specificity of approximately 82%. DISCUSSION
The focus of this study has been on the initial assessment of violence or suicide risk in a psychiatric ER and the extent to which such assessments are able to predict certain aspects of the subsequent course of hospitalization. It was found that patients who are immediately hospitalized after an initial ER evaluation are somewhat different from those who are discharged home. They tend to be younger, contain a higher proportion of males, are more likely to be single, have a more frequent history of substance abuse, are more likely to have made multiple suicide attempts in the past, and are more likely to present with suicidal ideation or gestures. All patients were assessed using an expanded lo-item version of the Bengelsdorf et al (1984) triage rating system. The results showed that the ER patients admitted to the inpatient wards scored significantly higher on every one of the items as well as on total score. They also scored higher on the frequency of both prior suicide attempts and violent episodes. These findings clearly indicate that the VASA scale significantly discriminates between admitted and discharged ER patients. As further support of this point, the sensitivity and specificity were both found to be quite high (82%) using a cutoff score of 11. It thus appears that the VASA scale may be a useful tool for clinicians in ERs to guide the clinical interview. The second key question of the study was to examine whether initial ER data have any predictive value in describing aspects of the course of short-term inpatient hospitalization. The results indicate that the number of prior suicide attempts is highly correlated with the likelihood of violent behavior in the hospital (r = .60). A history of substance abuse is also highly correlated with the likelihood of violent behavior in the hospital within 3 or 4 weeks after admission (r = .63). It was also found that three items of the VASA scale (lifetime history of suicide attempts, lack of social support systems, and inability to cooperate with the interviewer) correlate significantly with suicide risk in the hospital (r = .41) as estimated from nursing notes. It thus appears that some degree of prediction of acting out aggressive behavior is possible over relatively short time periods, using simple clinically obtained measures. We thus believe that the rating scales described here have value for the clinician to help in the identification of patients who need to be hospitalized, and may also serve as limited predictors of some aspects of hospital functioning. ACKNOWLEDGMENT The authors would like to thank Janie Lynn Feldman for assistance in data collection.
REFERENCES 1. Warner SL: Criteria for involuntary hospitalization of psychiatric patients in a public hospital. Ment Hygiene45:122-128,196l
VIOLENCE/SUICIDE
RISK ASSESSMENT
IN THE ER
343
2. Baxter S, Chodoroff B, Underhill R: Psychiatric emergencies dispositional determinants and the validity of the decision to admit. Am J Psychiatry 124:1542-1548, 1968 3. Bengelsdorf H, Levy LE, Emerson RL, et al: A crisis triage rating scale: Brief dispositional assessment of patients at risk for hospitalization. J Nerv Ment Dis 172:424-430, 1984 4. Zwerling I, Karasu TB, Plutchik R, et al: A comparison of voluntary and involuntary patients in a state hospital. Am J Orthopsychiatry 45:81-86, 1975 5. Skodal CH, Karasu TB: Emergency psychiatry and the assaltive patient. Am J Psychiatry 135:202-205, 1978 6. Browning CH, Tyson RL, Miller SI: A study of psychiatric emergencies. Part II. Suicide Psychiatry Med 2:359-366, 1971 7. Tarasoff V: Regents of the University of California et al: 13 1 Cal Rptr 14, 17 Cal 3d 425, 55 1 Pzd 334 (1976) 8. Rachlin S, Schwartz HI: Unforseeable liability for patients’ violent acts. Compr Psychiatry 371731-775, 1986 9. Feinstein R: Clinical guidelines for the assessment of immanent violence, in van Praag H, Plutchik R, Apter A (eds): Violence and Suicidality: Perspectives in Clinical and Psychobiological Research. New York, NY, Bruner Maze], 1990 10. Monahan J: The clinical prediction of violent behavior. US Department of Health and Human Services Publication No. ADM 8 l-921. Washington, DC, US Government Printing Office, 198 1