ORIGINAL ARTICLE

Risk Factors of Violence During a 4-Week Period in a Psychiatric Outpatient Population Alec Buchanan, PhD, MD,* Charla Nich, MS,* Kevin S. Douglas, LL.B., PhD,Þþ Theresa Babuscio, MA,* and Caroline J. Easton, PhD*

Abstract: The clinical impact of structured risk assessment instruments has been limited by a lack of information regarding a) their short-term accuracy and b) the relationship between change as measured by the instrument and a change in the risk for harm. Data were collected every 4 weeks on a) variables designed to resemble the items of a structured risk assessment instrument, b) substance use, c) social circumstances and mental state, and d) violent behavior. Scores on the variables designed to resemble the items of a risk assessment instrument were associated with violence during the ensuing 4 weeks. However, an increase in a subject’s score on these variables was not associated with violence. Instead, increasing cocaine use and increasing social conflict as described by the subject at interview were associated with violence during those weeks. Key Words: Violence, risk, outpatient treatment, substance abuse (J Nerv Ment Dis 2013;201: 1021Y1026)

A

ssessing the risk for violence to others is part of clinical management in psychiatry. The past 25 years have seen the development of a range of structured instruments designed to help clinicians assess risk (Quinsey et al., 2006; Webster and Hucker, 2007; Singh and Fazel, 2010; Singh et al., 2011, Singh et al., in press). Some of these are ‘‘actuarial,’’ designed to produce a score. Others use a ‘‘structured professional judgment’’ model, requiring the application of clinician discretion to classify a case as high, medium, or low risk. Both types of instrument have satisfactory psychometric qualities and have been shown to predict violence at levels of accuracy substantially better than those of chance (Mossman, 1994). The impact of these instruments on clinical practice has been limited by two concerns (Buchanan, 2008). First, the usual time at risk in studies demonstrating their predictive accuracy has been months or years, instead of the days or weeks during which risk assessments are often required. Second, despite suggestions from studies of criminal justice that falling scores predict reduced offending on probation (Raynor, 2007), it has not been shown that changes in scores on structured instruments are associated with changes in the risk for violence in mental health outpatient populations. However, it is change to which clinicians frequently wish to respond (Maden, 2007; Skeem and Monahan, 2011). Although the accuracy of structured instruments in predicting violence during the subsequent days and weeks has not been widely studied, there is some evidence from studies of hospital inpatients

*Department of Psychiatry, Yale University School of Medicine, New Haven, CT; †Department of Psychology, Simon Fraser University Burnaby, BC, Canada; and ‡Mid-Sweden University, Sundsvall, Sweden. Send reprint requests to Alec Buchanan, PhD, MD, Division of Law and Psychiatry, Department of Psychiatry, Yale University, 34 Park St, New Haven, CT 06519. E-mail: [email protected]. Copyright * 2013 by Lippincott Williams & Wilkins ISSN: 0022-3018/13/20112Y1021 DOI: 10.1097/NMD.0000000000000061

The Journal of Nervous and Mental Disease

that aspects of someone’s psychiatric condition are correlated with violence during these time intervals. A range of variables, including psychotic symptoms (McNiel and Binder, 1994) and structured risk assessment instrument items (McNiel et al., 2003), predict violent behavior in the weeks after acute psychiatric admission, for instance. Fewer data on the short-term correlates of violence are available from outpatient settings, in part because of the difficulty of studying risk factors when the base rate of violence is low. Those data that are available suggest that violence is commoner in the days after alcohol or multiple drug use (Mulvey et al., 2006) and when symptoms fluctuate rapidly (Odgers et al., 2009). Violence is commoner also in the week after subjects’ descriptions of high levels of anger but not in the week after their revealing a combination of thought insertion and persecutory ideas (‘‘threat control override’’ symptoms) or general psychological distress (Skeem et al., 2006). Risk variables that are susceptible to change are usually labeled ‘‘dynamic.’’ Using the Historical-Clinical-Risk ManagementY20 (HCR-20; Douglas et al., 2003; Webster et al., 1997) as a guide, a structured professional judgment instrument in widespread use, we generated a set of dynamic risk variables to test whether violence during the ensuing 4 weeks was associated, first, with scores on these variables or, second, with changes in the same scores. We then examined whether the associations we found were independent of substance use. Finally, we examined whether other known correlates of violence operate as risk factors when time at risk is limited to 4 weeks.

METHODS Subjects and Procedure The subjects were participants in a treatment study administered by the Forensic Drug Diversion Clinic of Yale University Department of Psychiatry. All had been referred by courts and criminal justice agencies. All had histories of domestic violence during the previous 12 months and of substance abuse. Of 90 men assessed and found eligible for the treatment study, 75 agreed to participate. All received a structured program of group and individual psychotherapy designed to treat substance abuse and reduce violence (Easton et al., 2007a). This study was approved by the Human Investigation Committee of Yale University (protocol no. 10492). The subjects were interviewed by a researcher at baseline and again at 4, 8, and 12 weeks. At each interview, the subject; the researcher; or, where required by the instrument, both completed a range of assessment instruments describing the subject’s psychological condition and rating the subject’s substance use and behavior during the past 28 days. In addition, the subject’s therapist completed assessments at 4, 8, and 12 weeks. Informants, typically the subject’s significant other, were interviewed by telephone regarding the subject’s behavior at the same time points. Substance use monitoring during this study was by self-report, urinalysis, and breathalyzer.

Variables The assessment instruments used all had satisfactory psychometric qualities. Those completed by the subject and the researcher

& Volume 201, Number 12, December 2013

www.jonmd.com

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

1021

The Journal of Nervous and Mental Disease

Buchanan et al.

& Volume 201, Number 12, December 2013

TABLE 1. Independent Variables

TABLE 1. (Continued)

Variables based on structured instrument items C1 Lack of insight Feeling others are to blame for troubles (BSI 4) I don’t have any problem that needs changing (URICA 1) C2 Negative Urges to beat, injure, or harm attitudes someone (BSI 40) Engaging in illegal activities for profit (ASI p14 q8) I am not the one with a problem (URICA 5) C3 Symptoms of Others controlling thoughts (BSI 3) major mental illness Thoughts of ending life (BSI 9) Hopelessness (BSI 35) Feeling seriously depressed (ASI p18q3[1]) Seeing or hearing things (ASI p18q3[3]) Suicide attempt (ASI p18q3[7]) C4 Impulsivity/ I feel irritated (STAXI 2) instability I feel angry (STAXI 3) I feel like yelling at somebody (STAXI 4) I feel like breaking things (STAXI 5) Feeling easily annoyed or irritated (BSI 6) Could not control temper outburst (BSI 13) C5 Unresponsive to Misses 950% of therapy appointments treatment/noncompliant (TSR 3) Late for appointments (TSR 4) Difficult to engage in therapy sessions (TSR 5) R1 Plans lack feasibility Therapist doubts what is being sought to achieve (WAI-T 12) Therapist and I have different ideas about the problem (WAI-C 27) Client and I have different ideas about the problem (WAI-T 27) R2 Exposure Not enrolled in school (ASI p3 q1) to destabilizers Not working (ASI p4 q7) Living with someone who has a drug problem (ASI p15 q6) R3 Lack of Dissatisfied with way time spent personal support (ASI p15 q8) R4 Potential problems Disagree that I have been doing something with compliance about the problems that have been or response bothering me (URICA 3) Disagree that worthwhile to work on my problem (URICA 4) Do not want to change anything about myself (URICA 8) Do not want help (URICA 20) Finds therapy confusing (WAI-C 7) Rated by therapist as finding therapy confusing (WAI-T 7) Says therapist does not understand what he wants (WAI-C 12) R5 Stress On probation or parole (ASI p13 q1) Arrested or charged in past month (ASI p13 q2) Awaiting charges, trial, or sentence (ASI p13 q5) Dissatisfied with marital status (ASI p15 q2) Dissatisfied with living arrangements (ASI p15 q5)

Circumstances and states of mind Conflicts Serious conflicts family/friends/others (ASI p16 q10) Irritability I fly off the handle (STAXI 16) Getting into frequent arguments (BSI 46) Suspiciousness Feeling that most people cannot be trusted (BSI 10) Feeling that people are unfriendly or dislike you (BSI 21) Fearfulness Suddenly scared for no reason (BSI 12) Feeling fearful (BSI 19) Substance use variables Dichotomous alcohol use Dichotomous cocaine use Dichotomous marijuana use Violence (outcome variable) Verbal violence Verbal abuse (CTS2 25) Shouting (CTS2 35) Threatened (CTS2 69) Acts Threw something (CTS2 7) Twisted arm or hair (CTS2 9) Pushed or shoved (CTS2 17) Punched or hit (CTS2 27) Slammed partner against wall (CTS2 37) Beat up (CTS2 43) Slapped (CTS2 53) Had unprotected sex (CTS2 15) Forced oral or anal sex (CTS2 19) Forced sex (CTS2 47) Threats for sex (CTS 2 57) Used knife or gun (CTS2 21) Choked (CTS2 33) Burned or scalded (CTS 2 61) Consequences Sprain, bruise, cut (CTS2 12) Victim unconscious (CTS2 24) Victim got medical attention (CTS2 32) Victim needed medical attention (CTS2 42) Victim broken bone (CTS2 56)

1022

www.jonmd.com

WAI-C indicates client section of the WAI; WAI-T, therapist section of the WAI.

were the Addiction Severity Index (ASI; McLellan et al., 1992), the Brief Symptom Inventory (BSI; Derogatis, 1993), the State-Trait Anger Expression Inventory (STAXI; Spielberger, 1996), the Revised Conflict Tactic Scale (CTS2; Straus et al., 1996), the client section of the Working Alliance Inventory (WAI; Horvath and Greenberg, 1989), and the University of Rhode Island Change Assessment (URICA; DiClemente and Hughes, 1990). At weeks 4, 8, and 12, the therapist completed the therapist section of the WAI and the Therapist Session Report (TSR; Orlinsky and Howard, 1975). Data from the assessment instruments were used to generate three groups of independent variables using the algorithm in Table 1. Data from items that did not relate to the questions we sought to * 2013 Lippincott Williams & Wilkins

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

The Journal of Nervous and Mental Disease

& Volume 201, Number 12, December 2013

Risk Factors of Violence

led to it being rated as present. Raters were blinded to independent variable scores when they were rating the subjects’ behavior.

answer were not used. The first group of independent variables comprised proxies of the 10 ‘‘risk management’’ and ‘‘clinical’’ items of the HCR-20. The HCR-20 variables are scored 0, 1, or 2. Each of the instrument items used to construct the proxy variables allowed either a negative or a positive response. We added the total number of positive responses on the items contributing to each proxy at each time point and allocated a score of 0, 2, or 1 to the subject according to whether the total of positive responses fell lower than the 33@ percentile for the sample as a whole, higher than the 66C percentile, or between the two. The ‘‘total score’’ at weeks 4, 8, and 12 was the mean of the 10 proxy variables. Three of the proxies, ‘‘unresponsiveness to treatment,’’ ‘‘plans lack feasibility,’’ and ‘‘potential problems with compliance,’’ were not available at baseline because the algorithm required data on treatment response to score them and the subjects had not yet started treatment. We prorated the baseline score to allow for this. The ‘‘total score’’ at baseline was, therefore, the mean of the other seven variables. The second group of independent variables related to substance use. Our subjects used only alcohol, cocaine, and marijuana regularly (see below). We assessed the subjects at interview on how many days they had used each of these during the past month. The third group of variables, circumstances and states of mind, contained known correlates of violence (Mullen, 1997) whose effectiveness as measures of short-term risk had not previously been studied. A variable in the third group was scored as present if a subject had a positive response on any of the items composing it. Violence occurring during the past 28 days was rated using CTS2 data and information obtained from collaterals using the algorithm in Table 1. Verbal violence was distinguished from physical aggression. Violence, whether verbal or physical, was rated as present if any of the items listed in Table 1 were endorsed. Where there was a difference between subject and collateral accounts, any endorsement of violence

Statistical Analysis The data were analyzed in two stages. The first stage assessed whether the subject’s score on the three groups of independent variables at weeks 0, 4, and 8 was associated with violence during the subsequent 4 weeks, that is, violence measured at weeks 4, 8, and 12, respectively. The second stage assessed whether the change in the subject’s score on an independent variable at weeks 4 and 8, measured by subtracting each subject’s score at weeks 0 and 4, respectively, from each subject’s score at weeks 4 and 8, respectively, was associated with violence in the 4 weeks subsequent to the change, that is, with violence measured at weeks 8 and 12, respectively. These relationships were examined using generalized estimating equations (GEEs). The GEE approach was used to account for intrasubject correlations across the three visits in which data were collected (Diggle et al., 2002). Each model specified an autoregressive correlation and included a within-subjects effect to accommodate the correlated data. The quantification of association was by log odds ratio (beta coefficient). The substance use variables were positively skewed, and for our analyses, we classed substance use as present or absent. The change variable for dichotomous independent variables was a threelevel predictor (j1, no use at time X with use at time Y; 0, no change in use; and 1, use at time X with no use at time Y). We ran discriminant function analyses (DFA) on the dichotomous violence outcomes using a combination of all of the independent variables. DFA allow for both a check on interdependence (by allowing all variables to be considered together) and a test of the strength of the association. Where suggested by our results and the findings of others,

TABLE 2. Association (Beta Coefficient) of Independent Variables With Violence Occurring in the Subsequent 4 Weeks Verbal Violence

Violent Acts

All Violence

b

Wald

p

b

Wald

p

b

Wald

P

178 165 175 186 121 89 180 180 186 192 174

0.24 0.07 0.19 0.54 0.09 0.89 0.08 0.20 0.92 0.50 0.66

0.79 0.08 0.13 6.93 0.08 2.52 0.05 0.11 2.05 0.36 5.09

0.37 0.77 0.53 0.01 0.78 0.83 0.74 0.15 0.55 0.02 0.02

1.72 0.35 0.08 0.86 0.31 0.90 0.02 0.08 0.05 0.96 0.44

8.13 0.79 0.02 6.38 0.18 0.89 0.00 0.91 0.92 0.28 0.31

0.00 0.37 0.89 0.01 0.67 0.35 0.97 0.34 0.34 0.60 0.58

0.54 0.18 0.24 0.67 0.13 0.73 0.01 0.06 0.37 0.97 1.01

3.91 0.54 0.62 9.77 0.16 1.93 0.00 0.01 1.37 2.19 1.14

0.05 0.45 0.43 0.00 0.69 0.17 0.99 0.92 0.24 0.14 0.29

191 191 191

0.05 0.08 1.73

0.04 0.03 8.35

0.85 0.87 0.00

0.02 0.89 1.52

0.02 0.89 1.52

0.98 0.20 0.02

0.08 0.39 1.67

0.07 0.72 7.48

0.79 0.39 0.01

111 186 188 188

0.82 0.58 0.68 0.69

1.83 3.38 4.25 4.50

0.18 0.07 0.04 0.03

0.30 1.35 1.49 1.20

0.08 5.16 4.70 3.65

0.79 0.02 0.03 0.06

0.71 0.57 1.02 0.83

1.41 3.19 9.26 6.39

0.24 0.07 0.00 0.01

www.jonmd.com

1023

Obs

Structured instrument item C1 Lack of insight C2 Negative attitudes C3 Symptoms of major mental illness C4 Impulsivity/instability C5 Unresponsive to treatment/noncompliant R1 Plans lack feasibility R2 Exposure to destabilizers R3 Lack of personal support R4 Potential problems with compliance or response R5 Stress Total score Substance use Alcohol Cocaine Marijuana Circumstances and states of mind Conflicts Irritability Suspiciousness Fearfulness

a

Values in bold are statistically significant at p G 0.05. a Obs = number of observations.

* 2013 Lippincott Williams & Wilkins

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

The Journal of Nervous and Mental Disease

Buchanan et al.

& Volume 201, Number 12, December 2013

TABLE 3. Association (Beta Coefficient) of Change in Independent Variables With Violence Occurring in the Subsequent 4 Weeks Verbal Violence

Violent Acts

All Violence

b

Wald

p

b

Wald

p

b

Wald

p

109 94 105 119 60 35 111 111 38 110 101

0.17 0.19 0.09 0.21 0.22 1.79 0.91 1.23 0.83 0.37 0.36

0.40 0.60 0.05 0.85 0.64 0.03 2.74 4.79 0.86 0.17 0.26

0.53 0.44 0.82 0.36 0.64 0.03 0.10 0.03 0.35 0.68 0.61

0.43 0.65 0.88 0.27 0.18 0.28 0.40 0.45 0.83 0.13 0.36

0.53 1.82 1.94 0.22 0.03 0.09 0.11 0.10 0.86 0.01 0.26

0.47 0.18 0.16 0.64 0.86 0.76 0.74 0.76 0.35 0.93 0.61

0.16 0.26 0.09 0.23 0.02 1.63 0.75 0.75 0.35 0.90 0.25

0.33 1.07 0.05 0.99 0.00 4.29 1.90 1.63 0.19 0.85 0.11

0.56 0.31 0.82 0.32 0.96 0.04 0.17 0.20 0.67 0.36 0.74

127 126 127

0.03 0.55 1.08

0.09 1.07 1.03

0.77 0.31 0.31

0.06 1.75 0.01

0.13 6.30 0.00

0.72 0.01 0.99

0.09 0.35 1.01

0.85 0.57 0.96

0.36 0.45 0.39

50 119 121 121

0.40 0.12 0.12 0.35

3.09 0.13 0.12 0.99

0.08 0.72 0.73 0.32

0.36 0.53 0.67 0.24

5.71 0.67 0.87 0.11

0.02 0.41 0.35 0.74

0.41 0.02 0.28 0.18

2.97 0.00 0.62 0.27

0.09 0.96 0.43 0.60

Obs

Structured instrument item C1 Lack of insight C2 Negative attitudes C3 Symptoms of major mental illness C4 Impulsivity/instability C5 Unresponsive to treatment/noncompliant R1 Plans lack feasibility R2 Exposure to destabilizers R3 Lack of personal support R4 Potential problems with compliance or response R5 Stress Total score Substance use Alcohol Cocaine Marijuana Circumstances and states of mind Conflicts Irritability Suspiciousness Fearfulness

a

Values in bold are statistically significant at p G 0.05. a Obs = number of observations.

we used the same statistical procedures to examine the association between substance use and violence occurring in the same month.

RESULTS The subjects’ mean age was 38 years. Seventy-five percent had graduated from high school, and 68% were in full-time employment. Using the Structured Clinical Interview for DSM-IV (First et al., 1995), all met the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), criteria for alcohol dependence. Three percent met the DSM-IV criteria for a lifetime diagnosis of depressive disorder; 8%, for a lifetime anxiety disorder; and 15%, for antisocial personality disorder (Easton et al., 2007b). Seventy remained in treatment at week 4; 66, at week 8; and 58, at week 12. At baseline, 49 (64%) of the subjects on whom data were available had engaged in any violence over the previous 28 days. 23 (30%) had engaged in violent acts over the same period. The equivalent figures at week 4 were 44% (any violence) and 9% (violent acts); at week 8, 32% (any violence) and 7% (violent acts); and at week 12, 21% (any violence) and 5% (violent acts). Forty-three percent tested positive for illicit drugs (cocaine, 34%; marijuana, 18%; opiates, G5%) during the study. Table 2 shows the associations between scores on the independent variables and violence during the subsequent 4 weeks. Of the variables designed as proxies for structured instrument items, ‘‘lack of insight’’ and ‘‘impulsivity/ instability’’ were associated with both violent acts and all violence. Marijuana use showed the same association, but alcohol and cocaine use did not. Scores on three of the four variables under ‘‘circumstances and states of mind,’’ irritability, suspiciousness, and fearfulness, were significantly associated with either violent acts (irritability and suspiciousness) or all violence (suspiciousness and fearfulness). Verbal violence was associated with structured instrument impulsivity, stress, and total score. 1024

www.jonmd.com

For all violence at week 12, when all of the variables listed here were included along with age and criminal history variables (age at first arrest, number of arrests, number of months in jail, and number of domestic violence arrests) in a stepwise DFA, two of the proxy variables, ‘‘impulsivity/instability’’ and ‘‘stress,’’ were significantly associated with violence and explained 61% of the variance. The model correctly classified 77% of the participants as nonviolent/violent (81% of the nonviolent and 64% of the violent; Wilks’ , = 0.31, W2 = 19.7, df = 2, p G 0.01). Table 3 shows the associations between changes in scores on the independent variables and violence during the subsequent 4 weeks. Of the variables designed as proxies for structured instrument items, changing scores on ‘‘plans lack feasibility’’ and ‘‘lack of personal support’’ were associated with verbal, but not physical, violence. Increasing cocaine use and an increased ability to identify conflicts with one’s peers and family were associated with physical violence. The associations with alcohol and cocaine use are smaller than we expected, given the known strength of the association between substance use and violence in psychiatric populations. We conducted one supplementary analysis, examining substance use in the same month as violence, rather than in the preceding month. GEEs for a variable comprising mean days of alcohol, cocaine, and marijuana use were significant for verbal violence (b = j0.18, Wald = 7.11, p G 0.01), violent acts (b = j0.22, Wald = 9.14, p G 0.01), and any violence (b = j0.19, Wald = 7.09, p = 0.01).

DISCUSSION Proxies of items in a structured risk assessment instrument are associated with verbal and physical violence during the subsequent 4 weeks. This has not previously been shown in an outpatient setting. The association is maintained when substance use is controlled for. Physical violence is associated also with marijuana use and with two of the four ‘‘circumstances and states of mind’’ * 2013 Lippincott Williams & Wilkins

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

The Journal of Nervous and Mental Disease

& Volume 201, Number 12, December 2013

variables studied here, irritability and suspiciousness. It is not, however, associated with rising scores on most of these variables. Instead, violent acts during the ensuing 4 weeks are associated with rising cocaine use and rising social conflict. Structured instruments may have a greater clinical role in assessing short-term risk in samples similar to this than has previously been thought. However, a rising score on most of the proxy variables was not associated with increased risk. Suggestions that scores on structured instruments can be used to track progress and intervene to prevent violence (Maden, 2007; Webster and Hucker, 2007) are not confirmed by these data and require further testing. We did not find the consistent and strong relationship that others have described between substance use and subsequent violence (Steadman et al., 1998). The supplementary analysis suggests an explanation that relates to the temporal relationship between substance use and violent behavior. The association between violence and substance use occurring in the same month is consistent with other reports of a short-term association between substance use and community violence (Mulvey et al., 2006). As in other samples, however, substance use in the same month could have occurred subsequent to an act of violence. The nature of the temporal association requires further investigation before causal inferences can be drawn. We found that all violence in the ensuing month was associated with a subjective sense of irritation (‘‘I feel irritated,’’ ‘‘I feel angry’’) contained in the impulsivity/instability proxy but not with the more objective measure of irritability that we used under ‘‘circumstances and states of mind.’’ This seems to speak to the importance of clinical evaluation in risk assessment. The data also confirm the clinical impression that talking to patients about conflicts in their domestic and social environments can reveal evidence that their risk for violence is changing. This study’s method generates several caveats. The subjects are not typical of clients in outpatient psychiatric services. All carried diagnoses of substance abuse, the base rate of violence was very high for an outpatient sample, and the levels of axis I symptoms were low. Although research has yet to demonstrate that the predictors of violence vary consistently with the clinical setting, inpatient data suggest that psychotic symptoms, for instance, are important when they are present (McNiel and Binder, 1994). Finally, we observed restricted variance in some of the change variables, which could have affected some of the associations with violence. Conclusions regarding the overall short-term predictability of violence in outpatient settings must await replication of these findings in different patient groups. With these qualifications, the method permitted evaluation of the relationship of subjective self-reported states and traits with subsequent violence in a clinically useful time frame. The variables that we examined addressed the range of factors that have been shown to be associated with short-term violence risk. The results derive from data collected using reliable instruments and multiple sources. Although no structured risk assessment instrument was used, the resources entailed in using instruments repeatedly mean that some operationalization of those items capable of change, whether using this approach or others, will be a necessary part of further defining the clinical role of structured instruments. The short-term correlates of violence in samples similar to this may fall into three groups. First, scores on variables that can be measured with structured instruments, lack of insight and subjectively defined feelings of irritability, predict violence during periods as short as several weeks. Changes in scores on these variables, however, do not seem to predict violence, at least in the circumstances described here. Instead, and second, violence during the ensuing several weeks is associated with rising scores on a different group of variables: social and domestic conflict and the use of some drugs. Although we did not include these in our proxy variables, it is likely that the changes these reflect can also be captured by structured instruments. Third, substance use is associated with all forms of violence, but the * 2013 Lippincott Williams & Wilkins

Risk Factors of Violence

time between use and subsequent violence is often so short as to make a patient’s self-report of even recent drug use of limited value as a predictor of violence during periods longer than a few days. We encourage researchers to investigate further a) which risk factors, if changed, might lead to changes in rates of violence; b) whether the method of measuring risk factors influences whether changes predict violence, and c) whether clinical intervention in response to increasing risk levels suppresses a statistical association between change and violence. ACKNOWLEDGMENT The authors acknowledge the contribution of Dr B. Rounsaville. DISCLOSURES This work was supported by the Donaghue Foundation (DF 0026) and by National Institute on Drug Abuse grants P50-DA0924 and K12 DA00167-11. Kevin Douglas is an author of the HCR-20. The authors declare no conflict of interest. REFERENCES Buchanan A (2008) Risk of violence by psychiatric patients: beyond the ‘‘actuarial versus clinical’’ assessment debate. Psychiatr Serv. 59:184Y190. Derogatis L (1993) Brief Symptom Inventory (BSI). Administration, scoring, and procedures manual (2nd ed). Baltimore: Clinical Psychometric Research. DiClemente C, Hughes S (1990) Stages of change profiles in outpatient alcoholism treatment. J Subst Abuse. 2:217Y235. Diggle P, Heagerty P, Liang K, Zeger S (2002) Analysis of longitudinal data (2nd ed). New York: Oxford University Press. Douglas K, Ogloff J, Hart S (2003) Evaluation of a model of violence risk assessment among forensic psychiatric patients. Psychiatr Serv. 54:1372Y1379. Easton C, Mandel D, Babuscio T, Rounsaville B, Carroll K (2007a) Differences in treatment outcome between male alcohol dependent offenders of domestic violence with and without positive drug screens. Addict Behav. 32:2151Y2163. Easton C, Mandel D, Hunkele K, Nich C, Rounsaville B, Carroll K (2007b) A cognitive behavioral therapy for alcohol-dependent domestic violence offenders: An integrated substance abuseYdomestic violence treatment approach (SADV). Am J Addict. 16:24Y31. First M, Spitzer R, Gibbon M, Williams J (1995) Structured Clinical Interview for DSM-IV. Patient edition. Washington, DC: American Psychiatric Press. Horvath A, Greenberg L (1989) Development and validation of the Working Alliance Inventory. J Couns Psychol. 36:223Y233. Maden A (2007) Treating violence: A guide to risk management in mental health. Oxford, England: Oxford University Press. McLellan T, Kushner H, Metzger D, Peters R, Smith I, Grissom G, Pettinati H, Argeriou M (1992) The fifth edition of the Addiction Severity Index. J Subst Abuse Treat. 9:199Y213. McNiel D, Binder R (1994) The relationship between acute psychiatric symptoms, diagnosis and short-term risk of violence. Hosp Community Psychiatry. 45:133Y137. McNiel D, Gregory A, Lam J, Binder R, Sullivan G (2003) Utility of decision support tools for assessing acute risk of violence. J Consult Clin Psychol. 71:945Y953. Mossman D (1994) Assessing predictions of violence: Being accurate about accuracy. J Consult Clin Psychol. 62:783Y792. Mullen P (1997) Assessing risk of interpersonal violence in the mentally ill. Adv Psychiatr Treat. 3:166Y173. Mulvey E, Odgers C, Skeem J, Gardner W, Schubert C, Lidz C (2006) Substance use and community violence: A test of the relation at the daily level. J Consult Clin Psychol. 74:743Y754. Odgers C, Mulvey E, Skeem J, Gardner W, Lidz C, Schubert C (2009) Capturing the ebb and flow of psychiatric symptoms with dynamical systems models. Am J Psychiatry. 166:575Y582. Orlinsky D, Howard K (1975) Varieties of psychotherapeutic experience. New York: Teachers College Press. Quinsey V, Harris G, Rice M, Cormier C (2006) Violent offenders. Appraising and managing risk (2nd ed). Washington, DC: American Psychological Association. Raynor R (2007) Risk and need assessment in British probation: The contribution of the LSI-R. Psychol Crime Law. 13:125Y138. Singh J, Fazel S (2010) Forensic risk assessment. A metareview. Crim Justice Behav. 37:965Y988.

www.jonmd.com

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

1025

The Journal of Nervous and Mental Disease

Buchanan et al.

Singh J, Fazel S, Gueorguieva R, Buchanan A (in press) Rates of violence by patients classified as high risk by structured risk assessment instruments. British J Psych. Singh J, Grann M, Fazel S (2011) A comparative study of violence risk assessment tools: A systematic review and metaregression analysis of 68 studies involving 25,980 participants. Clin Psychol Rev. 31:499Y513. Skeem J, Monahan J (2011) Current directions in violence risk assessment. Curr Dir Psychol Sci. 20:38Y42. Skeem J, Schubert C, Odgers C, Mulvey E, Gardner W, Lidz C (2006) Psychiatric symptoms and community violence among high-risk patients: A test of the relationship at the weekly level. J Consult Clin Psychol. 74:967Y979. Spielberger C (1996) State-Trait Anger Expression Inventory-Revised research edition. Professional manual. Odessa, FL: Psychological Assessment Resources Inc.

1026

www.jonmd.com

& Volume 201, Number 12, December 2013

Steadman H, Mulvey E, Monahan J, Robins P, Appelbaum P, Grisso T, Roth L, Silver E (1998) Violence by people discharged from acute psychiatric inpatient facilities and by others in the same neighborhoods. Arch Gen Psychiatry. 55: 393Y401. Straus M, Hamby S, Boney-McCoy S, Sugarman D (1996) The Revised Conflict Tactics Scales (CTS2): Development and preliminary psychometric data. J Fam Issues. 17:283Y316. Webster CD, Douglas KS, Eaves D, Hart SD (1997) HCR-20: Assessing risk for violence (Version 2). Burnaby, Canada: Mental Health, Law, & Policy Institute, Simon Fraser University. Webster C, Hucker S (2007) Violence risk: Assessment and management. Chichester, England: John Wiley & Sons.

* 2013 Lippincott Williams & Wilkins

Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Risk factors of violence during a 4-week period in a psychiatric outpatient population.

The clinical impact of structured risk assessment instruments has been limited by a lack of information regarding a) their short-term accuracy and b) ...
291KB Sizes 0 Downloads 0 Views