Psychometric evaluation of Chinese version of Violence Scale for objective rating among inpatients with schizophrenia

Download (0)



Psychometric evaluation of Chinese version of Violence Scale for

objective rating among inpatients with schizophrenia

Shing-Chia Chen and Hai-Gwo Hwu

Aim. The psychometric properties of the Chinese version of Violence Scale in clinical service setting were examined.

Background. Psychiatric inpatient’s aggressive act is a significant clinical issue in psychiatric service. A useful objective rating scale for prospective study and clinical application was mandatory.

Design. A prospective panel study.

Methods. The Chinese version of Violence Scale developed from Morrison’s Violence Scale. Sampled patients (n = 107) with schizophrenia spectrum, fulfilling the DSM-IV criteria, were recruited consecutively in a psychiatric acute ward of a university hospital over a period of one-year. The patients’ counts of the aggressive acts measured by the Chinese version of Violence Scale occurred in the past one-month prior to admission and in the prospective initial-week after admission were collected during their hospitalisation. The prospective occurrence was observed daily and summed at a fixed weekly point. The internal con-sistency, content validity and predictive validity of the Chinese version of Violence Scale were examined. Also, a confirmatory factors analysis by LISREL was conducted to examine its measurement structure.

Results. The Chinese version of Violence Scale follows a Poisson distribution of a fair quality (Cronbach’s a = 0Æ67). The Chinese version of Violence Scale with panel’s content validity has good predictive validity (r = 0Æ51, p < 0Æ001). Those correlated to one latent variable with six items, which constructs a core meaning of ‘threatening aggression toward others’. Conclusions. This panel study provides evidence for fair reliability and satisfactory validity of the Chinese version of Violence Scale. Internal consistency of the Chinese version of Violence Scale is limited and it may be because of the time-varying characteristic and hierarchical pattern of the behaviour items. To further investigate the count scale of the Chinese version of Violence Scale follows a Poisson distribution, the over-dispersion and weighting issues of aggressive acts were suggested to approach.

Relevance to clinical practice. This study highlights the measurement issues and implications of the Chinese version of Violence Scale for objectively rating psychiatric patients’ aggressive acts to further develop fitted nursing care and prevention program.

Key words: aggressive acts, behaviour observation, psychiatric nursing, psychometric evaluation, risk assessment, violence

Accepted for publication: 18 August 2008


The aggressive acts of psychiatric inpatients have been recognised as a significant clinical issue in psychiatric service

(Morrison 1988, Davis 1991, Gothelf et al. 1997, Hamolia 2001, Chen et al. 2005). The styles and prevalence of the aggressive acts were commonly related to those diagnosed of schizophrenia (Tam et al. 1996, Angermeyer 2000, Walsh

Authors: Shing-Chia Chen, PhD, RPN, Senior Lecturer, Department of Nursing, National Taiwan University College of Medicine and National Taiwan University Hospital, Taipei, Taiwan; Hai-Gwo Hwu, MD, Professor, Department of Psychiatry, National Taiwan University College of Medicine and National Taiwan University Hospital, Taipei, Taiwan

Correspondence: Shing-Chia Chen, Senior Lecturer, Department of Nursing, College of Medicine, National Taiwan University, 1, Jen-Ai Road, Sec. 1, Taipei 100, Taiwan, ROC. Telephone: +8862 23123456 (ext. 88433).


et al. 2004). But, the samples should exclude the dual diagnoses, such as personality disorder (Saverimuttu & Lowe 2000), substance abuse or dependence (Hoptman et al. 1999) or alcohol abuse (Walsh et al. 2004) to control the covariate of the results. The history of recent aggressive acts predicted the occurrence of aggressive acts (McNiel & Binder 1989, Walsh et al. 2004). In the first week of admission, the aggressive acts occurred relatively frequently (Tam et al. 1996).

In caring for patients’ aggression, a valid and reliable scale for clinical practice is important. The nursing process for caring psychiatric patient’s aggressive acts includes assessing the risk of aggression, establishing a care plan of management and prediction for aggression and then the monitoring and documentation of the process of intervention for patients’ aggressive acts (McNiel 1997). It is also important to measure aggressive acts for research purposes depending upon the questions to be addressed (Eron 1987). Previous studies were mostly limited to cross-sectional data obtained by retrospective design. Two commonly used measurement methods for assessing aggressive acts included self-report questionnaires and structured reviews on incident reports or medical charts. The first method suffers the limitation of subjectivity and social desirability. The latter tends to include the most noticeable acts of aggression only (Morrison 1988). Barratt and Slaughter (1998) and Kay et al. (1988) suggested use criterion measures for studying aggression in a longitu-dinal research design.

Among the previous four commonly used objective rating measures in aggression studies, the Violence Scale (VS) was chosen for this study because of its good sensitivity for measuring overt aggressive acts (Morrison 1993). In addition, the items of the VS had the most comprehensive behaviour descriptions and it provided typical examples for defining different categories and subcategories of aggressive acts in clinical observation. The other scales of the Staff Observation Aggression Scale (SOAS) (Palmstierna & Wistedt 1987) focused on the measurement for the incidents of aggressive acts; the Behavioral Rating Scale (BRS) (Cobb 1985), the Overt Aggression Scale (OAS) (Yudofsky et al. 1986)/ the Modified OAS (MOAS) (Kay et al. 1988) had only a summary of verbal or physical aggressive behaviours for measurement. In addition, the BRS and SOAS focused only on the outward aggressive acts. In clinical experiences and empirical studies, it appears that there is a close connection between the tendency to act aggressively outwards and inwards (Chen et al. 2005). Both of outward and inward aggressive acts are worthy to be measured and further investigated. Regarding another commonly used questionnaire HCR-20 developed to assess

risk for violence, it reflects risk factors related to the historical, clinical and risk management domain. Only the clinical subscale could be observed or inferred from patients’ current behaviours and the focus is on their psychological functioning (Webster et al. 1997, Douglas et al. 2003).

In exploring the primary construct and the generating theory (Bollen 1989), the exploratory factor analysis (EFA) demonstrated the construct validity of the VS meeting the criteria for loading aggressive acts toward property, other person and self (Morrison 1993). Furthermore, confirmatory factor analysis (CFA) was suggested to be more appropriate (Kline 1991) for examining the existing theory regarding the VS (Morrison 1993), but not well-done yet.

The aggression study usually had the shortcoming of difficultly in behaviour definition, retrospective research design, limited natural observation and underestimated chart documentation of aggressive acts, which need to be overcome (Turnbull et al. 1990, Chen 1997, Chen & Hwu 2002). A psychometric testing on a measure for rating aggressive acts objectively is mandatory for both research and clinical service. This study aims to develop the Chinese version of Violence Scale (VS-C) for objectively rating psychiatric patients’ aggressive acts clinically with sufficient validity.



This study is part of a prospective panel study for repeated measurement of patients’ aggressive acts during their hospi-talisation. This study applied the VS-C on the patient samples for measuring the aggressive acts occurred in the past one-month prior to admission and prospective initial-week after admission in an acute psychiatric ward.

Study participants

The patients (‡17 years of age) were consecutively admitted after completing the consent sheet for data collection with primary diagnosis of schizophrenia spectrum (SCZ) including paranoid, disorganised, undifferentiated and residual type of schizophrenia, schizophreniform disorder, schizoaffective disorder and schizotypal personality disorder. Patients fulfill-ing the diagnostic criteria of DSMIV (American Psychiatric Association 1994) were recruited for study from a 35-bed psychiatric acute ward of an urban university teaching hospital over a period of one year. The criteria of exclusion of the participants include comobidity with mental retarda-tion, antisocial personality disorder, organic brain


syndromes, and alcohol or substance abuse. If the diagnosis of SCZ was denied by the therapeutic teams in the hospital-isation course, these cases were excluded for study.


The theoretical background of the original VS was rooted in sociological and psychological domains including the concept of aggression (Bandura 1973), antisocial behaviour (Patterson 1982), aggression-frustration hypothesis (Dollard et al. 1939) and deviance (Hirschi 1969, Gibbs 1981). The scores of these randomly ordered 18 items of the VS were a Likert-type scaling based on frequency of aggressive acts. The scores ranged from 0–4 representing never, rarely, sometimes and frequently happening respectively. The internal consis-tency (a = 0Æ68 to 0Æ91), item analysis (all significantly different) and stability/reliability (r = 0Æ79) were satisfactory (Morrison 1993).

The Morrison’s VS was obtained and in agreement with original author for adapts. It was translated into Chinese version and was adapted for the convenience of clinical application and objective recording for patients’ occurrence of specific aggressive behaviour in the acute psychiatric ward and could be summated for further investigation. In considering different domains of aggressive acts with graded severity of the acts, the randomly ordered sequence of the items were arranged with the order from non-specific mild verbal to specific severe physical aggressive acts by three categories with counts scaling: (1) two items measured verbal threatening and three items measured physical damage toward ‘property’ (AggP), (2) four items measure verbal threatening and four items measured physical attack toward ‘other person’ (AggO) and (3) three items measured verbal idea expression and two items measured physical attempt toward ‘self ’ (AggS). It was blindly translated back into English and two bilingual persons with the expertise in English and Psychiatric Nursing exam-ined the equivalence between the two versions of scales and developed the final one with consensus. A panel of experts including board certified two psychiatrists and four senior psychiatric nurse professionals evaluated and judged the content validity of the VS-C being satisfactory.

Data collection

The data was collected by the nurse leader of the acute ward who had more than 20-years psychiatric nursing experience and taken major caring role of the patients. The patients’ basic data and their past one-month history of aggressive acts were collected using the VS-C in a face-to-face interview with patient and their primary caregiver relatives on admission. In

small cases, if the primary caregiver’s information could not be completed on admission, then we made an appointment with them to meet or interviewed them via phone call. If the counts data of patients’ past-month aggressive acts were not consistent between patients and their primary caregivers after the necessary clarification, the higher number was recorded. During admission, the nurse leader participated in the care and observation of any incident of aggression occurred throughout the day. Her participant observation was supple-mented by all other data sources including primary care nurses’ report, duty-shift report, records of ward-round team meetings and medical records. Based on all these observa-tions, the patients’ aggressive acts were rated using the VS-C and completed at her daily shift. She summated each patient’s daily counts data weekly on a Monday. The rating at the first weekly point of Monday after admission which was assumed to have the most high incident rate of aggressive acts was used as the initial-week data. The average length in days of the initial-week data after admission was close to one week (mean 5, SD 1Æ7 days).

Ethical consideration

This study has approved by the Human Subjects Review Committee in the Institutional Review Board of the Univer-sity Hospital. The written consents were obtained from the participants and their relatives. The participant’s basic rights were emphasised including personal privacy, treatment ben-efit and care quality. They also had the rights to refuse, remain silent or withdraw from the study. Confidentiality was maintained during the data collection and in the data entering procedures.

Statistical analysis

The consistency of Cronbach’s a and related criterion of the VS-C by Spearman’s correlation was analysed using the statistical software SPSS 12.0 (SPSS Inc., Chicago, IL, USA). The statistical software S-PLUS 6.0 (TIBCO Software Inc., Palo Alto, CA, USA) was used to figure the occurrence of each participant’s aggressive acts for reviewing the trend difference in individual level.

A CFA was conducted to identify the latent structure of the 18 items in the statistical software LISREL 8.43 (Scientific Software International Inc., Lincolnwood, IL, USA). The observed covariance matrix of the 18 items was the input. The model fitting process guided substantive knowledge using the following techniques: (1) A stepwise procedure for model selection was used to eliminate insignificant paths by the t-test (p-value > 0Æ05) backwardly and to add plausible


paths by the Modification Index (MI) forwardly. (2) The accuracy of the fit was assessed by checking whether the p-value of the chi-squared goodness-of-fit test exceeded 0Æ05 or, less strictly, the chi-squared values was less than two times of the degrees of freedom and (3) Poorly fitted points were located and examined. If the corresponding absolute value of the standardised residual was greater than 2Æ58, it was considered statistically significant at the 0Æ05 level in model diagnosis (Hair et al. 1998, p. 614-615). Finally, to validate the explanatory power of this fitted model, the score regression coefficients of this latent variable were applied to estimate the factor scores of each case for rechecking that the participants who got larger estimated factor scores really had greater aggressive acts in raw counts.


Participants’ demographic and clinical data

Of the 336 psychiatric patients admitted overall consecu-tively in the ward, 138 were eligible with primary diagnoses of SCZ. Of these 138 identified patients, 13 refused to participate and 18 were excluded for they were denied of the primary diagnosis of SCZ during their hospitalisation. The

remaining 107 participants were included in this study. Table 1 shows the participants’ demographic data.

The participants’ clinical types or subtypes of SCZ were paranoid (50Æ5%), disorganised (25Æ2%), undifferentiated (15Æ0%), residual (1Æ9%), schizoaffective disorder (3Æ7%) and schizotypal personality disorder (3Æ7%). There were 91 participants (85Æ0%) who had at least one past-month aggressive act and individual variation was big (mean 22Æ2, SD = 26Æ2, range 1–145 per week). After their admission to the ward, 63 participants (58Æ9%) had at least one initial-week aggressive act (mean 10, SD = 9Æ8, range 1–46). There were two items of the aggressive acts of VS-C, which never occurred during participants’ hospitalisation. These were (1) verbally threatened to harm another with a plan (vs-c 9) and (2) serious harm to self (vs-c 18). Compared with the aggressive acts toward property or other person, toward self were less frequently occurred (Table 2). Based on above results, the raw counts of occurred aggressive acts follow a Poisson distribution (i.e., mean is close to variance).

Internal consistency of the VS-C

The Cronbach’s a was computed to check the internal consistency of the VS-C. Excluding the two items without any incidents (vs-c 9 and vs-c 18), the value of the Cronbach’s a for the VS-C with 16 items was 0Æ67. Specifically, the values of the Cronbach’s a for the scales of AggP, AggO and AggS were 0Æ25, 0Æ63 and 0Æ33 respectively, whereas the verbal and physical subscales of AggP, AggO and AggS were (0Æ72, 0Æ26), (0Æ57, 0Æ55) and (0Æ42, 0Æ10) respectively.

Predictive validity of the VS-C

By the predicting criterion with past recent history of aggres-sive behaviour, the raw counts of the past-month aggresaggres-sive acts had statistically significant positive correlations with the initial-week aggressive acts (r = 0Æ51, p< 0Æ001), which showed a moderately well predictive validity of the VS.

Construct validity of the VS-C

After going through the stepwise model selection procedures, there were six items left in the final one-factor CFA model.

Table 1 Demographic data of the participants (n = 107)

Demographics n (%) Age (years) 17–20 10 (9Æ3) 21–30 41 (38Æ4) 31–40 33 (30Æ8) 41–50 9 (8Æ4) 51–60 12 (11Æ2) 61–69 2 (1Æ9) Gender Male 33 (30Æ8) Female 74 (69Æ2) Marital status Single 73 (68Æ2) Married 34 (31Æ8) Religious affiliation None 28 (26Æ2) Eastern 62 (57Æ9) Western 17 (15Æ9) Occupation No 79 (73Æ8) Yes 28 (26Æ2) Education background

2Junior high school 22 (20Æ6) Senior high school 44 (41Æ1)

3Junior college 41 (38Æ3)

Table 2 Frequency of each category of aggressive acts

Category n (%) Mean SD Range AggP 29 (27Æ1) 4Æ2 4Æ25 1–20 AggO 51 (47Æ7) 8Æ9 7Æ49 1–32 AggS 15 (14Æ0) 3Æ4 3Æ89 1–15


The key content of the six items included property used in a threatening manner without damage (vs-c 3), low-grade hostility (vs-c 6), loud verbal arguments (vs-c 7), verbally threatened to harm other person without a plan (vs-c 8), approached other person in a threatening manner without touching (vs-c 10) and low-grade physical/emotional harm to other person who need no medical care (vs-c 12). The variance–covariance matrix of the six items for CFA model-ling input was shown at Table 3. The Chi-squared goodness-of-fit test indicated that the fit of this CFA model to the observed data was acceptable (v2(8)= 10Æ45, p = 0Æ23)

(Fig. 1). Other goodness-of-fit indices included Root Mean Square Error of Approximation (RMSEA) of 0Æ05, Normed Fit Index (NFI) of 0Æ94, Non-Normed Fit Index (NNFI) of 0Æ96, Goodness of Fit Index (GFI) of 0Æ97 and Adjusted Goodness of Fit Index (AGFI) of 0Æ92. All the absolute values of the standardised residuals of the one-factor CFA model were less than 2Æ58.

The coefficient of determination (R2) of each structural

equation provides a statistical measure of reliability (Bollen

1989). The order of the R2-values for the six items in the final CFA model were vs-c 6 (0Æ95), vs-c 7 (0Æ47), vs-c 3 (0Æ47), vs-c 10 (0Æ37), vs-c 12 (0Æ06) and vs-c 8 (0Æ06). They were in the same order as the values of the factor loadings in the one-factor CFA model. As most of the indicators were aggressive acts towards other persons (vs-c 6, vs-c 7 and vs-c 8 belong to the subscale of the verbal AggO; and vs-c 10 and vs-c 12 belong to the physical AggO) and towards property (vs-c 3 was of the physical AggP), this single factor could be labelled as ‘threatening aggressive acts towards others’.


This study presents that the Chinese version of Violence Scale (VS-C) had a satisfactory degree of psychometric property and it is recommend as a potentially useful measurement scale in aggression study or clinical service of psychiatric inpatients. This VS-C had moderate reliability of internal consistency with Cronbach’s a of 0Æ67. The content, predic-tive and construct validity are shown to be satisfactory.

For the purpose of descriptive evaluation of aggressive acts, this study performed two modifications on the original version of VS. The first is to replace the originally global severity by frequency rating with behaviour count of each item of aggressive act, which already has the connotation of severity. The aggressive acts were measured commonly with the Likert-type scales in a cross-sectional study in the past (Morrison 1988, 1993, Gothelf et al. 1997). In the clinical setting, longitudinal study is necessary to understand more

Table 3 The variance-covariance matrix of the core items

vs-c 3 vs-c 6 vs-c 7 vs-c 8 vs-c 10 vs-c 12 vs-c 3 4Æ00 vs-c 6 4Æ08 9Æ08 vs-c 7 2Æ89 5Æ76 8Æ38 vs-c 8 0Æ12 0Æ41 0Æ38 0Æ27 vs-c 10 1Æ07 1Æ28 1Æ79 0Æ06 1Æ85 vs-c 12 0Æ01 0Æ41 0Æ33 0Æ01 0Æ13 0Æ35 2·11 FACTOR 1 vs-c 3 0·47 4·41 0·25 1·17 0·33 -1·14 1·00 2= 10·45, χ df = 8, p-value = 0·23486, RMSEA = 0·054 2·93 1·99 0·15 vs-c 6 vs-c 7 vs-c 8 vs-c 10 vs-c 12 1·38 0·12 0·83

Figure 1 The path diagram of the one-factor CFA model from LISREL.


patients’ health-related behaviours (Huang et al. 2006) in the disease and/or treatment course. To obtain more quantitative information for research and clinical purpose, the ‘count’ scale was used in the VS-C instead. The VS-C could obtain raw counts data of each behaviour item, the categories (property, other person and self) and subcategories (verbal and physical) of aggressive acts and can be summated (aggressive acts outwards and inwards) for further investiga-tions. Thus, our data and the analysis results were unique and more informative about the aggressive acts. The second is to rearrange the behaviour items into ordered three categories of aggressive acts toward property (AggP), other person (AggO) and self (AggS). Each category contains two subcategories of verbal and physical domain. The order of the aggressive act items was arranged by the potential severity of the behaviour items. This modification will facilitate the behaviour record-ing of the specific aggressive acts, which is clearly observed in the clinical service ward. It seems unnecessary to arrange these behaviour items in a random order, as it is not a subjective rating scale susceptible to social desirability. However, the modified count scaling of the VS-C may induce a potential statistical problem because of the over-dispersion of count data among individuals (Lindsey 1998). It is usually assumed that count outcomes being rare events follow a Poisson distribution (Kutner et al. 2005) and non-parametric statistics were adapted to be used (Selvin 1995). As many participants had zero counts in some of the aggressive acts and very few had larger counts, the frequency distributions of the count are very likely skewed to the right. We suggest the count data obtained by the VS-C may have to take a log transformation (after adding 0Æ1 to each count to avoid the log0 problem) or other suitable power transformation (i.e., Yq, q < 1) on the count to make its distribution more

symmetric (Hamilton 1992) and then treat it as a continuous measurement for further statistical operation.

The overall internal consistency of the VS-C with 16 items was fair (a = 0Æ67) as suggested by Polit and Beck (2004) in group-level comparisons and Hair et al. (1998) in exploratory research. However, the scales of AggP, AggO and AggS and each of those verbal or physical subscales have a relatively low internal consistency. This may suggest that the items are not equally loaded on the same construct (i.e., tau-equivalence is violated) (Huang et al. 2006) or need an adjustment in a group (Rattray & Jones 2007) for individual aggressive acts. This result was generally not consistent with that of Morrison’s result (Morrison 1993) from the original VS. This may be because of the difference in the scaling methods – the actual counts vs. the ordinal Likert scale. Morrison (1992) had also found that there was a hierarchy of aggressive acts toward other person in the part of VS. That is,

the items of each category of the VS as ordered by the severity of the aggressive acts on a unidimensional pattern actually had a property like a Guttman scale (Morrison 1993). As a causal indicator effect, this kind of structure in the VS-C led to low internal consistency and affected negatively the results of the factor analysis (Bollen 1989). We suggest that ‘weighting’ approach may be a worthwhile approach to deal with the hierarchy of aggressive acts in further study. For example, to integrate quantity and severity, one might be willing to set equivalence between two mild occurrences with an item and one occurrence of a severer item.

The inter-rater reliability was usually suggested for a scale with relatively long-lasting attribution of items (Huang et al. 2005, Rattray & Jones 2007). For the time-varying charac-teristics of the aggressive acts, it was impossible to do such an investigation, although not impossible, to measure the inter-rater reliability of the VS-C by two nursing staffs at the same time in the setting treating varied and acute clinical status.

Validity is the extent to which the indicators accurately measure what they are supposed to measure. In this study, translation, back translation and an expert panel evaluation to assure the content validity of the VS-C. This validity assessment assures the cross-cultural equivalence including content, semantic, technical, criterion and conceptual equiv-alence (Flaherty et al. 1988), especially for a behaviour rating scale of descriptive items usually with less complicated cultural metaphor (Brislin 1986, Jones & Kay 1992). However, measuring an act as aggressive involves an assess-ment of intent, which is a subjective process likely to be influenced by cultural issues. In this study, the data collector not only observed but also taken part in the caring process of patients’ aggressive acts. It was more sensitive to assess patients’ intent by recognising patients’ cultural origins and unique perceptions related to their aggression.

We approached the predictive validity of the VS-C, based on a history of aggressive acts, which was usually considered to be the most important predicting factor of aggression (Tardiff et al. 1997, Walsh et al. 2004). The finding of this study supports this assertion that the past one-month aggressive acts predict the subsequent aggressive acts after admission. In this study, the past aggressive history measured in a one-month period prior to admission is short enough to cover the covariates of fluctuations of aggressive acts to have more accurate recall, which was consistent by Morrison’s survey (Morrison 1998) who recalled and rated the amount of psychiatric nursing staff’s encountered aggression for the past month to collect data.

We examined the construct validity of the non-parametric data of VS-C using the CFA approach (Lo 1996, Hair et al. 1998). It is beyond Morrison’s conduct of EFA (Morrison


1993), which explores the initial factor model of VS to seek a few underlying dimensions that account for the patterns of co-variation among the aggressive acts only (Hamilton 1992). In our study, unidimensionality was confirmed by the acceptable fit of the single-factor model (Hair et al. 1998). Thus, it was valid to use factor analysis and provide estimates of values of our study. Accordingly, a shorter version of the VS-C can measure that dimension of the aggressive acts in clinical service setting. It can be applied to assess inpatients’ proneness of general aggression in psychiatric acute ward for identifying patients at high risk for aggression, providing specific nursing care and building up a suitable prevention program.

Schizophrenia spectrum is considered to be one risk factor for aggression. This study is based on the fact that the patients of schizophrenia spectrum represent the largest proportion of the inpatient psychiatric service. In this study, there are 58Æ9% of inpatients with schizophrenia spectrum who had aggressive acts during the initial-week interval of study period. The data are close to the past findings about patients’ aggressive acts which occurred in the risk-peak initial week after admission (Kay et al. 1988, Chen 1997). This provides useful information for clinical service of the psychiatric inpatient. However, further studies are needed to validate the VS-C in various psychiatric diagnostic groups.


This panel study provides evidence for fair reliability and satisfactory validity of the Chinese version of Violence Scale (VS-C) and recommends the VS-C to be useful for clinical psychiatric service and research purposes. Internal consis-tency of the VS-C is limited to be grouped and it may be because of the time-varying characteristic and hierarchical pattern of the behaviour items. The aggressive acts occurred in one month prior to admission are measured by the VS-C, which has good predictive value of the prospective occur-rence of aggressive acts in the initial week after admission. Confirmatory factor analysis validates the one identified factor used for measuring the aggressive acts. We suggest that further studies are desirable to investigate the count scale of the VS-C follows a Poisson distribution in approaching the over-dispersion and weighting issues of aggressive acts.


This study was supported by grants from the National Science Council (NSC90- B- 002- 327; NSC91- 2314-B002- 328), Taiwan, ROC. We specially thank Dr Fu-Chang Hu for dedicating his statistical expertise to our study. We

appreciate Dr Eillen F Morrison for kindly providing us with the original Violence Scale and insightful discussions, the data assistance from Jui-Yuan Huang and the statistical assistance from Su-Chen Chen.


Study design: SCC, HGH; data collection and analysis: SCC and manuscript preparation: SCC, HGH.


American Psychiatric Association (1994) Diagnostic and Statistical Manual of Mental Disorder: DSM IV, 4th edn. APA, Washington, DC.

Angermeyer MC (2000) Schizophrenia and violence. Acta Psychia-trica Scandinavica 102(Suppl. 407), 63–67.

Bandura A (1973) Aggression: A Social Learning Analysis. Prentice-Hall, NJ.

Barratt ES & Slaughter L (1998) Defining, measuring, and predicting impulsive aggression: a heuristic model. Behavioral Sciences and the Law 16, 285–302.

Bollen KA (1989) Measurement models: Appendix 6A LISREL pro-gram. In Structural Equations with Latent Variables. John Wiley & Sons, NY, pp. 223.

Brislin RW (1986) The wording and translation of research instru-ments. In Field Methods in Cross-cultural Research (Lonner WJ & Berry JW eds). Sage Publications, Beverly Hills, CA, pp. 137–164. Chen SC (1997) Study of patients’ violent behavior in an acute psychiatric ward. The Journal of Nursing Research 5, 195–205. Chen SC & Hwu HG (2002) Aggressiveness in psychotic patients:

psychopathological theory and clinical management. Taiwanese Journal of Psychiatry 16, 261–275.

Chen SC, Hwu HG & Williams RA (2005) Psychiatric nurses’ cog-nition and anxiety in managing psychiatric patients’ aggression. Archives of Psychiatric Nursing 19, 141–149.

Cobb BA (1985) A descriptive correlational study exploring the relationship between adult psychiatric patient role strain and vio-lence in an inpatient setting. Dissertation Abstracts International 45, 2496B.

Davis S (1991) Violence by psychiatric inpatients: a review. Hospital and Community Psychiatry 42, 585–590.

Dollard J, Miller NE, Doob LW, Mowrer OH & Sears RR (1939) Frustration and Aggression. Yale University Press, New Haven CT. Douglas KS, Ogloff JR & Hart SD (2003) Evaluation of a model of violence risk assessment among forensic psychiatric patients. Psychiatric Services 54, 1372–1379.

Eron LD (1987) The development of aggressive behavior from the perspective of a developing behaviorism. American Psychologist 42, 435–442.

Flaherty JA, Gaviria FM, Pathak D, Mitchell T, Wintrob R, Richman JA & Birz S (1988) Developing instruments for cross-cultural psychiatric research. The Journal of Nervous and Mental Disease 176, 257–263.

Gibbs JP (1981) Norms, Deviance and Social Control: Conceptual Matters. Elsevier, New York.


Gothelf D, Apter A & van Praag HM (1997) Measurement of aggression in psychiatric patients. Psychiatry Research 71, 83–95. Hair JF, Anderson RE, Tatham RL & Black WC (1998) Structural equation modeling. In Multivariate Data Analysis, 5th edn. Prentice-Hall, NJ, pp. 614–615.

Hamilton LC (1992) Principal components and factor analysis. In Regression with Graphics: A Second Course in Applied Statistics. Wadsworth, Belmont, CA, pp. 249–288.

Hamolia CD (2001) Preventing and managing aggressive behavior. In Principle and Practice of Psychiatric Nursing, 7th edn (Gail WS & Michele TL eds). Mosby, St. Louis, Mo, pp. 635–657.

Hirschi T (1969) Causes of Delinquency. University of California Press, CA, Berkeley.

Hoptman MJ, Yates KF, Patalinjug MB, Wack RC & Convit A (1999) Clinical prediction of assaultive behavior among male psychiatric patients at a maximum-security forensic facility. Psychiatric Service 50, 1461–1466.

Huang MC, Chen CH, Yu CY, Yuh JF, Lai YC & Lin SK (2005) Reliability of the Chinese version of the revised Clinical Institute Withdrawal Assessment for Alcohol scale. Taiwanese Journal of Psychiatry 19, 250–253.

Huang CL, Lin HH & Wang HH (2006) Psychometric evaluation of the Chinese version of the Fagerstrom Tolerance Questionnaire as a measure of cigarette dependence. Journal of Advanced Nursing 55, 596–603.

Jones EG & Kay M (1992) Instrumentation in cross-cultural research. Nursing Research 41, 186–188.

Kay SR, Wolkenfeld F & Murrill LM (1988) Profiles of aggression among psychiatric patients: I. Nature and prevalence. The Journal of Nervous and Mental Disease 176, 539–546.

Kline RB (1991) Latent variable path analysis in clinical research: a beginner’s tour guide. Journal of Clinical Psychology 47, 471–483. Kutner MH, Nachtsheim CJ, Neter J & Li W (2005) Logistic regression, Poisson regression, and generalized linear models. In Applied Linear Statistical Models. McGraw-Hill, NY, pp. 618. Lindsey JK (1998) Counts and times to events. Statistics in Medicine

17, 1745–1751.

Lo SK (1996) Factor analysis: a discussion of practical problems commonly encountered in nursing research. Chang Gung Nursing 7, 70–75.

McNiel DE (1997) Correlates of violence in psychotic patients. Psychiatric Annals 27, 683–690.

McNiel DE & Binder RL (1989) Relationship between preadmission threats and later violent behavior by acute psychiatric inpatients. Hospital and Community Psychiatry 40, 605–608.

Morrison EF (1988) Instrumentation issues in the measurement of violence in psychiatric inpatients. Issues in Mental Health Nursing 9, 9–16.

Morrison EF (1992) A hierarchy of aggressive and violent behaviors among psychiatric inpatients. Hospital and Community Psychiatry 43, 505–506.

Morrison EF (1993) The measurement of aggression and violence in hospitalized psychiatric patients. International Journal of Nursing Studies 30, 51–64.

Morrison EF (1998) The culture of caregiving and aggression in psychiatric settings. Archives of Psychiatric Nursing 12, 21–31. Palmstierna T & Wistedt B (1987) Staff observation aggression scale,

SOAS: presentation and evaluation. Acta Psychiatrica Scandinav-ica 76, 657–663.

Patterson GR (1982) A Social Learning Approach: Coercive Family Processes. Castalia Press, Eugene OR.

Polit DF & Beck CT (2004) Nursing research: Principles and meth-ods, 7th edn. Lippincott Williams & Wilkins, Philadelphia, PA, pp. 421, 498, 500.

Rattray J & Jones MC (2007) Essential elements of questionnaire design and development. Journal of Clinical Nursing 16, 234–243. Saverimuttu A & Lowe T (2000) Aggressive incidents on a

psychi-atric intensive care unit. Nursing Standard 14, 33–36.

Selvin S (1995) Poisson regression analysis. In Practical Biostatistical Method. Belmont, Wadsworth, CA, pp. 455–496.

Tam E, Engelsmann F & Fugere R (1996) Patterns of violent inci-dents by patients in a general hospital psychiatric facility. Psychi-atric Services 47, 86–88.

Tardiff K, Marzuk P, Leon A & Portera L (1997) A prospective study of violence by psychiatric patients after hospital discharge. Psychiatric Services 48, 678–681.

Turnbull J, Aitken I, Black L & Patterson B (1990) Turn it around: short-term management for aggression and anger. Journal of Psychosocial Nursing and Mental Health Services 28, 7–10. Walsh E, Gilvarry C, Samele C, Harvey K, Manley C, Tattan T,

Tyrer P, Creed F, Murray R & Fahy T (2004) Predicting violence in schizophrenia: a prospective study. Schizophrenia Research 67, 247–252.

Webster CD, Douglas KS, Eaves D & Hart SD (1997) HCR-20: assessing risk for violence (Version 2). Vancouver, Mental Health, Law, and Policy Institute, Simon Fraser University, Psychological Assessment Resources Inc., Canada.

Yudofsky SC, Silver JM & Jackson W (1986) The overt aggression scale for the objective rating of verbal and physical aggression. American Journal of Psychiatry 143, 35–39.


Table 1 shows the participants’ demographic data.

Table 1

shows the participants’ demographic data. p.4
Table 1 Demographic data of the participants (n = 107)

Table 1

Demographic data of the participants (n = 107) p.4
Figure 1 The path diagram of the one- one-factor CFA model from LISREL.

Figure 1

The path diagram of the one- one-factor CFA model from LISREL. p.5