The psychiatric department at St. Olavs University Hospital, Trondheim, Norway has a catchment area of 140.000 inhabitants. About 700 patients above 18 years with acute psychiatric conditions are admitted each year. Norwegian acute psychiatric emergency services are publicly funded and available to everyone. All the patients in the catchment area are admitted to this department. Acute admissions to other psychiatric hospitals occur only if inhabitants temporarily reside outside the catchment area when the need for acute admittance arises.
The acute department consists of two ordinary closed emergency wards each with a PICU area with 4 beds. The patients were admitted to the acute ward with most free capacity. One ward was used for the study, and the patients excluded from the study were admitted to the other ward.
The physician on duty evaluated all the patients acutely admitted to the ward. Patients assessed to be in need of PICU were admitted to the PICU area and included in the study except those with dementia, mental retardation or autism to a severe degree, and patients not speaking Norwegian or English. They were excluded at evaluation before entering the PICU area and admitted to the other ward.
The study ward consists of an ordinary closed acute ward area (310 m2) and a PICU area (190 m2). The main entrance leads to the ordinary area of the ward. In the end of the corridor a locked door separates the PICU area from the ordinary area. The PICU area consists of two wings with two single patient rooms in each. The patients stay mostly in the wings together with nurses, and contact with other patients is limited. The PICU area thus provides segregation from other persons. A sketch of the ward is published previously .
Data from the present study stems from two different inclusion periods in the PICU with background data published previously [15, 16]. The two samples were comparable for all measurements at inclusion . In the first inclusion period the entrance door to the PICU area was permanently locked and the doors inside the PICU leading to the wings were kept permanently closed (inclusion1). In the second period the entrance door to the PICU area was removed and the doors inside the PICU were permanently open (inclusion 2). These two conditions made it possible to compare two different methods of organizing a PICU . Inclusion 1 was a completely segregated PICU-condition, while inclusion 2 was a condition giving the patients opportunity to choose between a certain level of isolation and calmness by staying in the patient room/wing of the PICU or move to the main part of the ward ensuing exposure to other patients, more staff-members and increased amounts of sensory and emotional stimuli. The clinical staffs were similar in these inclusion periods. There was no significant difference in occupation of PICU-beds between inclusion 1 and 2.
The environment-related condition in the inclusions thus was the use of the PICU as a segregation area or not . The difference between the inclusions as a possible predictive factor for SOAS-R incidents is the parameter "Segregation".
Symptoms, general psychopathology, function and behaviour were assessed with a single rating on the first day in the PICU with the Positive and Negative Syndrome Scale (PANSS) for schizophrenia  with time criterion the last 24 hours, the Global Assessment Scale Split version (GAF-S), and the Brøset Violence Checklist (BVC) . GAF-S is based on DSM-4's GAF  and is a two-item scale measuring global symptoms (GAF-S-Symptoms) and functioning (GAF-S-Function) separately. BVC is a six-item observer-rated scale scoring behaviours that predicts imminent violence in psychiatric inpatients [20, 21]. BVC predictions are traditionally performed three times daily [20, 21] as opposed to the present study evaluating predictive properties for the next three days from a single assessment at admittance. Since psychometric properties of The PANSS used in an emergency setting with time criterion last 24 hours is not previously tested, two trained ward nurses evaluated this in a separate pre-study. Through scorings of 3 video-taped patient interviews  and assessments of 12 consecutively admitted acute emergency patients, the ward nurses demonstrated excellent inter-rater reliability for total PANSS sum, sums of positive (Pearson's r = 0.96), negative (r = 0.84) and general subscales (r = 0.87), as well as the 30 single items.
Violent or threatening incidents were recorded with the Staff Observation Scale-Revised (SOAS-R) [23, 24]. The SOAS-R severity score ranges from 0-22 points with higher scores indicating greater severity . A SOAS-R score ≥ 9 indicates a serious incident [12, 24].
Therapeutic- and control steps taken and nurses' observations were coded on a 23-item checklist. These therapeutic steps and observations included for instance "frequency of testing out and pushing limits", "intensity of testing out and pushing limits", "need to set limits", all prescribed medication, side effects, formal restrictions (restrictions regarding visits and telephone), staff contact time, use of newspapers, and visits from relatives. Depending on the type of item each were scored on scales 0-4 (0 = not present, 4 = very much) or 0-1 (0 = not used, 1 = used). Specially trained unit nurses did all the ratings. The first rating with the items "frequency of testing out and pushing limits", "intensity of testing out and pushing limits", and "need to set limits" was used as a possible predictor after an initial, short observation of the patient's behaviour at admittance and right after entrance to the PICU in order to evaluate whether the experienced staff's assessments of the patients' immediate behaviour could predict SOAS-R incidents for the rest of the study period.
At admittance the physician on duty evaluated the patients' need for PICU on a scale with scorings 1-4 (1 representing no need to 4 representing absolute need). Four categories of reasons for admittance to PICU were noted (1: Patient's own wish, 2: Need of close observation from diagnostic or medical reasons, 3: Reduction of stimuli, or 4: Control of behaviour). If more than one reason was present, the physician indicated the dominating category. "Physician's prediction" is an index defined by giving the patients with category 4 reason for admittance the scorings on "patients' need" of PICU, and the rest of the patients value 0. "Physicians prediction" therefore has scorings 0-4 with increasing value indicating increasing assumed probability for violent or threatening incidents.
The patients were systematically examined for substance use at admittance, at evaluation with ward psychiatrist the first weekday after admittance, and at discharge from PICU. In the first period (November 13 2000 to March 25 2001) (n = 56), urine samples were analysed on clinical suspicion of substance use. In the second period (October 1 2001 to March 21 2002) (n = 62), all admitted patients had urine- and blood samples taken within a few hours after admission.
Diagnoses according to ICD-10 Diagnostic criteria for research  were set by consensus in the department's staff, including at least three specialists in psychiatry of whom at least two personally had examined the patient.
All data were analysed using the Statistical Package for the Social Sciences (version 11.0). Demographic and clinical variables were described using means and frequencies. Independent t-tests were used to compare differences between groups. Multiple logistic regression forced-entries were performed to examine the extent of the predictor variables' associations with SOAS-R incidents. The variable SOAS-R was categorised as incidents and non-incidents. Relative risk (RR) was calculated with Fischer's exact test and a generalized mixed model with Poisson distribution of SOAS-R incidents. The missing values were replaced with the mean score for each item. Statistical significance was defined as a two-tailed p < 0.05.
The study was approved by "The Regional Medical Research Ethics Committee, Central Norway."