Paper
John Olav Roaldset, Aalesund Hospital, More and Romsdal Health Trust. Centre for Research and Education in Forensic Psychiatry, Oslo University Hospital. The Norwegian University of Science and Technology, Norway
Stål Bjørkly, Centre for Research and Education in Forensic Psychiatry, Oslo University Hospital. Molde University College, Norway
Keywords: Biomarker, risk assessment, violence
Introduction
The formal launch of the bio-psycho-social model (BPSM) of medicine dates back to the seventies [1]. The mechanisms behind violent behavior among persons with psychiatric disorders are complex and involve psychological, social and biological factors. While a substantial part of risk assessment research has combined psychological and social factors, neurobiological factors have for the most part been investigated without involving or even controlling for the impact of the two other main groups of risk factors [2]. Even if numerous theoretical models of violence have been introduced, most of these models fail to integrate a tripolar understanding. Recently, however, some authors have paidmore attention to bio-psycho-social approaches to violence in mentally disordered people with focus on both theoretical [3] and empirical [4]
research. The basic assumption of the BPSM is that the underlying factors and behavioural manifestations of violence are the result of neurobiological, individual, environmental and socio-economical causes.
However, (i) The specificity of the proposed underlying factors for understanding the development of violence is unclear, (ii) Underlying factors are often assumed to reflect specific aspects of neurobiological functioning. However they are very often monitored indirectly by using behavioural rating scales, (iii) The complexity of the interaction between BPS factors is augmented by the fact that they all are subject to change. These arguments points in two compatible directions: To continue the work on breaking complex underlying factors into well-defined components to better understand their association with violence, and to investigate possible interactions between underlying factors that increase the risk of violence. This requires adequate and reliable construction of multivariate prediction models [5].
The multivariate model approach is based on the assumption that model variables should not only be evaluated in isolation for their predictive validity, but rather on their added (incremental) predictive validity beyond existing or established predictors. To further recognize the role of the new factor in the model one has to assess whether it is a correlate, risk factor or causal factor. Correlates are simply variables that are associated with for instance violence, but they are generally not predictive or causal. Risk factors are correlates that are shown to predict another variable (e.g. violence). To show that a variable is a risk factor it has to precede the outcome (violence). Causal risk factors can change and when they do so they impact the risk for violence. A few studies have indicated biological measures as risk factors, e.g. low central concentration of serotonin metabolites. Results from a growing number of investigations have identified biological variables as correlates or biomarkers of violence. An example of this is low values of total cholesterol (TC). Several cross- sectional studies have identified a significant correlation between low TC and violence [6-8]. In most hypotheses the effect of low cholesterol on aggression is linked to decreased serotonin in the central nervous system caused by reduced serotonin transport through cell membranes [9]. Therefore, it is important to emphasize that TC is treated as a correlate and not as a risk factor. A biomarker may be of a rapid or a slowly fluctuating type, but it is very unlikely that it will remain completely stable and unchangeable. Measures of biomarkers may be obtained by standard hospital procedures (e.g. routine blood samples), or by specialized medical examination that is not part of standard hospital procedures (e.g. fMRI).
Patients own perceptions of risk for future violence have been sparely examined in the literature, but results from two recent studies of patients from (i) forensic psychiatry [10] and (ii) acute psychiatry [11]
indicated that patients’ self-report of risk had acceptable validity in predicting severe violence within the first two months after discharge from forensic hospitals, and in predicting any violence within the first three months after discharge from an acute psychiatric ward.
There have been notable advances over the last three decades in violence risk assessment of mentally disordered persons. In spite of this the ROC-AUCs of current instruments lie around .70 [12] This finding indicates the need for assessment of additional risk markers and factors. In this research we combined a biological marker and self-reported risk of violence with a ten-item screen. The Violence risk screening- 10 (V-RISK-10) is a structured screen with good reliability and predictive validity [13, 14].
The main aim of this investigation was to test different BPS models for violence risk assessment. Baseline variables (V-RISK-10, patients own perception of risk for future violence (SRS), and lipid levels) were rated or measured at admission, and compared with violent behavior recorded during hospital stay and the first year after discharge.
Method
Setting and subjects
The study was conducted at the acute psychiatric ward in Ålesund Hospital and included all acute admitted patients during one year (n=489) who gave written consent to participate in the study (n=254). Some patients were lost during follow-up and some had incomplete forms so the final study population consisted of 134 patients. Among follow-ups (n=134) there was significantly more patients who were involuntary admitted, more patients with bipolar disorders, and longer median hospital stays compared to the missing patients (n=355).
Measures
Total cholesterol (TC)
TC, high density lipoprotein (HDL), low density lipoprotein (LDL) and triglycerides were measured in blood drawn between 08 and 09 am within the first three days after admission, with a fasting period from 12 pm the day before.
SRSOur literature search failed to show any empirical research on patients’ self-reported “direct” opinion of subsequent violent behavior. Due to lack of other available instruments, a two-item self-report screen (SRS) with a seven-point scale was constructed to measure the patients’ judgments of their subsequent risk for violence. The patients were asked to respond to two questions pertaining to two timeframes:
For the time you are staying in the ward/for the first three months after discharge from the ward; what is your opinion about the risk that you: Will threaten other people by acting violently? Will act violently against others? For each question, the patients choose one of the seven respond options to express their risk estimate: no risk (will definitely not happen), low risk (will hardly happen), moderate risk (limited to certain situations), high risk (in many situations), very high risk (almost permanent risk), don’t know the risk, and, will not answer about the risk [11].
V-RISK-10
TheV-RISK-10 [15] is a10-itemchecklist developed for acute psychiatry [16, 17]. The items are: 1. Violent threats, 2. Violent acts, 3. Substance abuse, 4. Major mental illness, 5. Personality disorders, 6. Lack of insight, 7. Suspiciousness, 8. Lack of empathy, 9. Unrealistic planning, and 10. Future stress-situations.
Each item has a brief scoring instruction presented in the scoring form. Items are scored on a three-point scale (Table 1): 0 (No – The item definitely is absent or does not apply), 1 (Maybe/moderate – The item is possibly present, or is present only to a limited extent), 2 (Yes – The item is definitely present), Omit (Don’t know –There is insufficient valid information to permit scoring the item).
The V-RISK-10 was constructed to include factors predictive of later violence. It is a structured clinical screen with historical, clinical and future risk assessment items. This means that it is a clinical guide to early detection of possible violence risk and not an actuarial instrument. The latter contains fixed and explicit algorithms to estimate the specific probability or absolute likelihood that a person will engage in violence in the future. The forerunner of the V-RISK-10, the Ps33, was inspired by the HCR-20, but the present 10-item version is substantially different from the Ps33. For instance, four of the items tap information concerning both past and present identification of the individual risk factor.
Outcome measures
During hospital stay, violence was monitored by the Report Form for Aggressive Episodes (REFA) [18, 19].
Shortly after the occurrence of an episode in the ward, information about the time and the characteristics of the situation, the precipitating factors, the persons involved, and the severity of the episode, were recorded.
The inpatient episodes were recorded as threats or acts. Violent episodes that occurred after discharge were recorded in scoring schemes. Data from criminal and police records concerning violent threats and
acts included convictions, charges, and withdrawal of charges for violent crime by reason of insanity were combined with hospital data into a common outcome variable. The episodes were categorized into threats, less severe acts, and severe acts with operational definitions of each category. The scores on each category were: No, not present, Yes, present, or Don’t know if present.
Statistical analysis
Data were analysed using SPSS version 16.0. Bivariate and multivariate logistic regression analysis was used to compare the TC, SRS, and single items and total-scores of V-RISK-10. We used a backward procedure in the multivariate single variable analysis. Exp (B) was used as odds ratio for occurred episodes.
Results
TC was a significant predictor of violent behaviour during follow-up in all bivariate analysis of the ten single items of V-RISK-10 (p-values between 0.005 – 0.001). The same goes for the ten single items and SRS (p-values between 0.030 - 0.001). TC and SRS were both significant in bivariate analysis of these two variables (p-values = 0.001 and 0.011, respectively). The SRS was not significant in bivariate analysis with V-RISK-10 score (p-values = 0.345 and <0.001, respectively), while both TC and V-RISK-10 total-score were significant in bivariate analysis with these two variables (p= 0.010 and <0.001, respectively).
All ten single items of V-RISK-10, the SRS and TC were entered into a backward logistic regression analysis to obtain the best prediction model. Results showed that only TC (OR=0.27, 95% CI=0.13-0.54, p<0.001), V1 (violent acts, OR= 4.8, 95%CI=1.5-15, p=0.008), V3 (substance abuse, OR=4.2, 95%CI=1.3-13, p=0.014, and V10 (future stress situations, OR=4.8, 95%CI=1.4-17, p=0.015) remained significant.
Discussion and conclusion
We report results from a comparison of total cholesterol and patient self-report (SRS), and also when total cholesterol and SRS values were compared to single items of the V-RISK-10. Low cholesterol was a significant predictor of inpatient and outpatient violent behavior during the first 3-4 months after blood sampling at admission. SRS was also a significant predictor of violence at 3-4 months follow-up, which also was the case for V-RISK-10. The multifaceted risk assessment model yielded a significant increase in explained variance beyond that of the V-RISK-10 alone, and in multivariate analyses TC significantly accounted for variance beyond that of SRS and V-RISK-10. The association between low cholesterol and aggression has been linked to impulsive aggression but not to predatory aggression [20-22]. A recent predictive validity study of the HCR-20 showed that prediction of future violent events was particularly complicated in disorders characterized by impulsive behaviour [23] The significant contribution of TC as a biomarker for violent behaviour indicates that risk assessment of persons with impulsive violent behaviour may be enhanced by including biological markers
However, the results of our investigation must be interpreted with caution due to the small sample size.
Other limitations are: (i) A high number of staff recorded the outcome measures, and episodes may have been underreported, (ii) Due to the follow-up procedures, violence from patients discharged into community may have run undetected more often than in patients followed up by psychiatric services, (iii) Consenting patients had lower rates of post-discharge violence than the non-consenting sample, and (iv) Differences between the subsamples indicate that non-consenters were characterized by more severe illness and a lack of insight. Taken together these differences may limit the external validity of the results.
Future research may want to involve large-scale prospective study designs, other candidate biomarkers, and more advanced assessment of psycho-social factors. At last but not at least we suggest that epigenetic research may be of paramount significance to further development of reliable and valid bio-psycho-social models for risk assessment of violence.
References
1. Engell, G.L., The need for a new model: A challenge for biomedicine. Science 1977. 196: p. 129-137.
2. Loeber, R. and D. Pardini, Neurobiology and the development of violence: Common assumptions and controversies.
Philosophical Transactions of the Royal Society B: Biological Sciences, 2008. 363: p. 2491-2503.
3. Steinert, T. and R. Whittington, A bio-psycho-social model of violence related to mental health problems. International Journal of Law and Psychiatry, 2013. 36(0): p. 168-175.
4. Roaldset, J.O., et al., A multifaceted model for risk assessment of violent behaviour in acutely admitted psychiatric patients.
Psychiatry Research, 2012. 200(2–3): p. 773-778.
5. Moons, K.G.M., et al., Risk prediction models: I. Development, internal validation, and assessing the incremental value of a new (bio)marker. Heart, 2012. 98(9): p. 683-690.
6. Boscarino, J.A., P.M. Erlich, and S.N. Hoffman, Low serum cholesterol and external-cause mortality: Potential implications for research and surveillance. Journal of Psychiatric Research, 2009. 43(9): p. 848-854.
7. Golomb, B.A., Cholesterol and violence: is there a connection?[see comment]. Annals of Internal Medicine, 1998. 128(6):
p. 478-87.
8. Hillbrand, M. and R.T. Spitz, Cholesterol and aggression. Aggression and Violent Behavior, 1999. 4(3): p. 359-370.
9. Engelberg, H., Low serum cholesterol and suicide. The Lancet, 1992. 339(8795): p. 727-729.
10. Skeem, J.L., et al., The Utility of Patients’ Self-Perceptions of Violence Risk: Consider Asking the Person Who May Know Best. Psychiatric Services, 2013. 64(5): p. 410-415.
11. Roaldset, J.O. and S. Bjørkly, Patients’ own statements of their future risk for violent and self-harm behaviour: A prospective inpatient and post-discharge follow-up study in an acute psychiatric unit. Psychiatry Research, 2010. 178(1): p. 153-159.
12. Fazel, S., et al., Use of risk assessment instruments to predict violence and antisocial behaviour in 73 samples involving 24 827 people: systematic review and meta-analysis. BMJ (Clinical research ed.), 2012. 345: p. pp e4692.
13. Hartvig, P., et al., The first step in the validation of a new screen for violence risk in acute psychiatry: The inpatient context.
European Psychiatry, 2011. 26(2): p. 92-99.
14. Roaldset, J.O., P. Hartvig, and S. Bjørkly, V-RISK-10: Validation of a screen for risk of violence after discharge from acute psychiatry. European Psychiatry, 2011. 26(2): p. 85-91.
15. Hartvig, P., et al. Violence Risk Screening 10 (V-RISK-10). 2007.
16. Hartvig, P., et al., Brief checklists for assessing violence risk among patients discharged from acute psychiatric facilities: a preliminary study. Nordic Journal of Psychiatry, 2006. 60(3): p. 243-8.
17. Bjørkly, S., et al., Development of a brief screen for violence risk (V-RISK-10) in acute and general psychiatry: An introduction with emphasis on findings from a naturalistic test of interrater reliability. European Psychiatry, 2009. 24(6):
p. 388-94.
18. Bjorkly, S., Report form for aggressive episodes: Preliminary report. Perceptual and Motor Skills, 1996. 83(3, Pt 2): p.
1139-1152.
19. Bjorkly, S., A ten-year prospective study of aggression in a special secure unit for dangerous patients. Scandinavian Journal of Psychology, 1999. 40(1): p. 57-63.
20. Conklin, S.M. and M.S. Stanford, Premeditated aggression is associated with serum cholesterol in abstinent drug and alcohol dependent men. Psychiatry Research, 2008. 157(1-3): p. 283-7.
21. Troisi, A., Low cholesterol is a risk factor for attentional impulsivity in patients with mood symptoms. Psychiatry Research, 2011. 188(1): p. 83-87.
22. Vevera, J., et al., Cholesterol concentrations in violent and non-violent women suicide attempters. European Psychiatry, 2003. 18(1): p. 23-27.
23. Gray, N.S., J. Taylor, and R.J. Snowden, Predicting violence using structured professional judgment in patients with different mental and behavioral disorders. Psychiatry Research, 2011. 187(1-2): p. 248-253.
Correspondence
John Olav Roaldset johnolr@gmail.com