Short-term prediction of threatening and violent behaviour in an Acute Psychiatric Intensive Care Unit based on patient and environment characteristics

Threatening and violent behaviour by psychiatric inpatients are major concerns in psychiatric practice [1, 2]. Aggression has negative consequences for patients and staff. Some studies indicate an increasing frequency [3, 4]. Reduction of severity and incidence of threatening and violent incidents are important in order to improve quality of care in psychiatric facilities. Prediction of violence is consequently important in order to initiate preventive measures. Risk factors and accuracy of predictions in different psychiatric settings have been extensively described [5].

In Psychiatric Intensive Care Units (PICUs) and emergency services violent incidents are frequent and short-term predictions of violence important [6]. Patients in these settings present in severe crisis often complicated by behavioural dyscontrol, substance use, and multiple axis 1 diagnoses [7]. Prediction and prevention of aggression in emergency services patients could thus potentially be based on numerous data drawn from individual medical and social history, treatment conditions, behaviours and psychopathology. A history of violence in medical records, and a diagnosis of schizophrenia or substance abuse are known predictors for violence [5, 8]. However, in the emergency setting with acute admittances of unknown patients these factors are often unknown and thereby clinically irrelevant. These are indications that such predictors have more limited value in emergency clinical settings compared to community or forensic settings [5, 8–10]. In acute settings short-term prediction may be more accurate based on global clinical judgment [9–11], or rating scales specifically developed to assess short-term prediction of aggressive incidents [12]. In a retrospective study from a PICU population Bjørkdahl et al demonstrated that nurses can predict violence to a high degree for the next 24 hours with three times daily assessments using the Brøset Violence Checklist (BVC) [12].

Differences in methods, designs, outcome measures and settings make it difficult to compare studies of short-term prediction of violence. The weaknesses in most designs are underreporting of incidents, lack of correction for therapeutic interventions, and a disregard of the influence from the treatment environment [5, 9]. Environment-related conditions like extent of segregation of behaviourally disturbed patients are believed to influence rate of aggressive incidents [13].

Rating scales specifically developed to measure frequency and severity of violent or threatening incidents, a prospective research design, and control of environment-related conditions and treatment factors are believed to minimize the problems in research methods.

The frequency of threatening and violent incidents differs with the stages of psychiatric disorders [14]. The highest frequencies of such incidents are reported in acute stages of in-patient hospital stay which indicate the first few days in emergency settings.

The aims of the present prospective study were to investigate clinically relevant patient and environment-related predictive factors for threatening and violent incidents the first three days in a psychiatric acute emergency population based on evaluations at admittance. These patient-related conditions are actuarial data (age, gender, admission status and diagnosis), assessments of symptoms, function and behaviour, therapeutic interventions the first day, global clinical impression assessed by physician on duty, global clinical impression assessed by treating nurses, and predictive properties from a single BVC rating at admittance; while the environment-related condition is the use of the PICU as a segregation area or not.

The present study gives evidence for a positive effect of segregation quantitatively determined in relation to other variables. We are not aware of the presence of such evidence so far. This finding is highly relevant for both clinical practice and the design of psychiatric wards. Effects of ward space and architecture are sparsely studied. Palmstierna et al found that patients with schizophrenia were more likely to be aggressive in a crowded ward [31]. In a second study the same authors did not find a decline in the frequency of aggression in spite of a reduction of the number of beds by 50% [32]. Nijman et al were unable to document a decline in aggressive incidents after extending space in a ward [33]. In the present study the number of beds and the space were identical in the two inclusions [15]. This finding indicates that an important factor in reducing aggressive incidents is the separation of single patients or patient groups in the ward, not the physical space in terms of square meters per patient.

Our data predicting short-term violent and aggressive incidents in a PICU are in accordance with previous studies from acute wards. Generally the predictive values from actuarial data are limited. The global clinical evaluation "Physicians prediction" from physician on duty, and the observer-rated scale scoring behaviours predicting imminent violence in psychiatric inpatients (BVC), were more suitable for predicting short-term violent and threatening incidents in the PICU setting. Since BVC is based on observer rated scorings of behaviour, the present study demonstrates that experienced staff members in acute settings are able to globally predict short-term violence in their patient populations.

We found no association between SOAS-R ratings and psychopathology measured by PANSS total, PANSS subscales, and GAF-S. This finding is similar to Swett et al's [34]. Steinert et al. in contrast found that scorings on the seven-item PANSS-positive scale correlated significantly with the number of threatening or aggressive incidents in a sample of acutely admitted in-patients [35]. Findings from studies using the Brief Psychiatric Rating Scale (BPRS) [36] or PANSS are contradictory. Using the full scale PANSS is time consuming, but this systematic questioning discloses important aspects of symptoms and makes the staff able to take these into account in therapy. This may lower the number of violent or threatening incidents, and make conclusions from different studies difficult [16].

The observer rated instrument BVC is based on reports of the most frequent behaviours observed prior to violent incidents, and it assesses the presence or absence of the six behavioural states confusion, irritability, boisterous behaviour, verbal threatening, physical threatening, and attacking objects [20]. It has demonstrated satisfactory properties in forensic and acute settings [21, 37], and now in a PICU setting. The instrument is short, practical and easy to administer in routine care. Systematic uses of standardised instruments like BVC give staff opportunities to focus on preventive measures towards limited numbers of high-risk patients.

Involuntary admission status did not predict SOAS-R incidents in the present study. This finding is contrary to Nijman et al.'s who found a history of involuntary admission to be a predictor of aggressive behaviour [10]. This may partly be due to different criteria for involuntary admissions. Some countries (e.g. Dutch law [10]) allow forced hospitalisation only when a patient's behaviour constitutes a direct and clear danger to the patient or others. Norwegian law extends this concept to allow involuntary admissions in other cases of severe mental illness based on the need of treatment.

Diagnoses of substance use or schizophrenia are reported to be predictive factors for aggressive incidents [5, 28–30]. This was not supported by the present data. Assessment of substance use is difficult and under-reporting is a problem. In the present study substance abuse was extensively assessed. Our study indicates that presence of substance use diagnoses do not facilitate threatening and violent behaviour among patients in a PICU setting [38].

Several studies with different interventions have been conducted to assess the effects of preventive measures on aggressive incidents [39]. Conclusions are difficult to draw due to shortcomings in the research designs like lack of control conditions, possible under-reporting of aggressive incidents and staffs' awareness of their wards being objects of research. There are also indications that systematic monitoring of aggressive incidents with for instance SOAS-R, increases the staffs' awareness of risk factors eventually leading to a decrease in numbers of incidents. Nijman et al [39] compared the effects of several possible aggressive incidents reducing interventions in a closed psychiatric admission ward with two similar control wards. The main results were a significant reduction of aggressive incidents in all the three wards. The reduction in the intervention ward and control wards were 62% and 43%, a difference that turned out to be non-significant. The present study indicates that global experience in staff and structured instruments may help identify patients where preventive measures should be considered. These measures should include physical separation of these high-risk patients from the others.

This study has weaknesses. "Physicians prediction" is an index composed of the physician on duty's global impression of the patients need and reason for admittance to PICU. This is not a validated instrument, but reflects the main outcome of what goes on in the mind of the experienced clinician. The nurse-rated 23-item checklist "Therapeutic- and control steps taken and nurses' observations" has similar shortcomings. The SOAS-R incidents are few, but comparable to other studies. The mean severity score of the incidents is moderate.

The study sample is a consecutive, highly-selected sample of acutely admitted patients assessed in a PICU. The methods and facilities used for emergency psychiatry differ between countries [40]. This special PICU have similarities to what Bowers names "open area seclusion" [40]. Different selections in different facilities may give other results. However, the principles of stimulus reduction and segregation from other patients are similar to other segregation settings. This may indicate that our study generally illustrates the effects seen also in traditional seclusion settings.

Segregation of patients raises ethical and legal questions. During the last years there have emerged new legislations, recommendations, court cases and professional guidelines to control the use of coercive measures in psychiatry. The recurring message in all of these guidelines is the need to practice caution when applying segregation in the form of seclusion [41]. The present segregation setting represents an effective alternative to seclusion with the patients staying mostly in the wings together with nurses [15].

The strengths of the study are numerous. First of all this is a prospective design in a naturalistic patient population from a defined catchment area. We have used robust validated instruments assessing symptoms, general psychopathology, function, behaviours and end-point measure. The routine screening for substance abuse was comprehensive. Therapeutic and control steps taken the first day were assessed. The degree of potential under-reporting of aggressive or threatening incidents is limited due to the prospective design and daily prompting for registrations. Finally the influence of the physical environment is incorporated in evaluation.

In the psychiatric emergency setting the procedure of segregation of single patients predicts lowered numbers of short-term aggressive incidents. Predictions at admittance should further be based on a global clinical evaluation from the physician at admittance, or rating with instruments designed to predict short-term aggression in psychiatric in-patients like the Brøset Violence Checklist.