The aim of this study is to describe the frequency and severity of aggressive incidents in acute psychiatric wards in the German speaking part of Switzerland. Secondary aims are to estimate the extent of underreporting and to explore associations of patients' characteristics with the occurrence of aggressive incidents. The study design aims to meet some of the methodological problems of existing studies.

After approaching all 87 acute psychiatric wards in the study area 24 (27.6%) wards from 12 hospitals agreed to participate in the study.

Before embarking on this study a survey of all wards within the study area using a questionnaire covering data on the size of the wards, staffing, the facilities for managing aggression and violence 36 was conducted. Additionally, staff nurses were asked to rate the severity of the problem and the resources for aggression management.

Aggressive incidents were recorded by the revised Staff Observation Aggression Scale (SOAS-R) 37. The SOAS-R covers provoking factors, the means used by the patient, the target of aggression, the consequence for the target, and the measures to terminate the aggression. The scale is to be completed by staff members witnessing aggressive behaviour of a patient whereby aggression is defined as any verbal, non-verbal, or physical behaviour that was threatening (to self, others, or property), or physical behaviour that actually did harm (to self, others, or property) 38. The severity of the incidents was measured using the SOAS-R-scoring system which ranges from 0 to 22 points, and additionally by a 100-mm-Visual Analogue Scale (VAS) with the endpoints "not severe incident" and "very severe incident".

Data on patients were taken from the hospitals data bases and included dates of admission and discharge, age, gender, and main psychiatric diagnosis according to ICD-10.

In order to estimate possible underreporting visits to most of the study wards were conducted at randomly selected dates. We investigated the written shift reports of the previous three days and noted all descriptions of aggressive incidents. After termination of the study period, we controlled these incident data of the respective wards for SOAS-R-data covering these events, and estimated the severity of the incidents in cases where SOAS-R-forms were missing.

We included all aggressive incidents directed towards other persons or objects, but excluded purely auto-aggressive incidents. As recommended by the authors of the SOAS-R, events with a severity of 9 or more points on the SOAS-R-severity score were regarded as severe incidents (Henk Nijman, personal communication). Physical attacks were defined as recordings on the SOAS-R fulfilling the two criteria 1) the means of aggression = objects OR dangerous objects OR parts of the body AND 2) the target of the aggression = a person other than the patient her- or himself. We calculated incidence rates for all aggressive incidents, for severe incidents, for physical attacks, physical aggression and purely verbally aggressive incidents. Incident rates were expressed as the proportion of patients involved in aggressive events, as event-rates per 100 hospitalisation days with 95%-confidence-intervals, as event rates per bed per annum, and as the proportion of days with one or more event. In cases of multiple treatment episodes of patients in the study period, we used the first episode as index episode for the patient related analyses.

We employed binary logistic regression to test for associations between patients' characteristics and the incidence. We used the presence or absence of at least one more severe aggressive incident as binary outcome measure. We calculated crude odds ratios (OR) for each of the independent variables separately.

Statistical significance was determined using the traditional cut-off level α = 0.05. We used SPSS (Version 10.0, SPSS, Chicago, IL) and CIA (Confidence Interval Analysis, Version 2.1, University of Southampton, UK) for statistical analysis.

Information on ward characteristics were obtained from 82 of the 87 wards the study area. Twenty four wards with a total of 388 beds participated, representing 12 out of the 32 hospitals and 25% of all acute wards and 28% of all acute beds in the German speaking region of Switzerland. The number of beds per ward ranges from 9 to 19 with a mean number of beds of 16; the mean staff-patient-ratio was .77 Full Time Equivalents (FTE) per bed. Fifteen (62%) of the wards are always closed, 9 wards part of the time, and 22 of the 24 wards have at least one seclusion room. According to the ratings of the staff nurses aggression is deemed a large or very large problem on 13 (54%), as a small or medium problem on 11 (46%) wards. None of the ward nurses rated aggression as very small or no problem at all. Seven (29%) of the wards regarded their resources for managing aggression as insufficient. No significant differences were observed between participating and non-participating wards regarding the number of beds and staff-patient-ratios. The participating wards include a higher proportion of closed wards and of wards having seclusion rooms, wards rating aggression as a problem and the rating their resources for aggression management as insufficient (table 1).

The observation period was three month per ward. The study period covered 72 month (24 wards × 3 month) and included all seasons of the year: 13 month of observation in January-March, 5 in April-June, 16 in July-September, and 38 in October – December.

During the study period 2017 patients (46.6% females, mean age 39.4 ± 13.9 years, range 12 – 96 years) accounted for 2344 treatment episodes on the 24 study wards. The hospitalisations comprised of 60.9% voluntary and 39.1% involuntary admissions giving rise to a total of 41'560 treatment days. The length of the treatment episodes ranged between one and 133 days (median 9, Mean 17.7 ± 22.0 days). The patient's ICD-10 diagnoses comprised of schizophrenia, schizo-type and delusional disorders (ICD F2: 29.3%), mood (affective) disorders (F3: 16.7%), mental and behavioural disorders due to psychoactive substance use (F1: 24.9%), neurosis (F4: 13.7%) and personality disorders (F6: 4.1%), and other ICD-10 categories (F0, F5, F7, F8, F9, others: 3.4%) (4.7% missing).

A total of 760 aggressive incidents were reported, including 396 incidents with a SOAS-R severity score of 9 or more and 252 physical attacks (the latter do not necessarily score 9 or more). 157 incidents were purely verbal aggression. These figures correspond to an overall incidence rate of 1.829 (95% CI 1.701 – 1.963) aggressive incidents per 100 treatment days, 0.950 incidents with a severity score of ≥ 9 (95% CI 0.859 – 1.049), and 0.606 (95% CI 0.534 – 0.686) attacks against persons, respectively (table 2). The incident rates per bed per year were 7.08 for all incidents, 3.32 for more severe incidents (SOAS-R ≥ 9 points) and 2.35 for physical attacks. On 14.2% (one in 7 days, one per week) of the calendar days within the study period at least one aggressive incident was registered, on 10.0% (1 in 10 days) of the days at least incident with a severity of ≥ 9 and in 7.9% of the days at least one physical attack (one in 13 days, one every two weeks).

The severity of all 760 incidents on the SOAS-R severity scale ranged from 0 to 21 with a mean severity of 8.80 ± 4.88 and a median of 9 points. On the VAS, the mean severity rating was 34.40 ± 26.26 with a median of 28 and a range from 0 to 100. There is a significant correlation between these two severity scores of 0.321 (Spearman Rho, p < .001). The correlations were 0.352 in incidents of purely verbal aggression and 0.269 in incidents of physical aggression (table 3).

The mean severity of physically aggressive incidents was significantly higher than the severity of purely verbally aggression (Mann-Whitney-U Test p < 0.001 for SOAS-R severity score and p = 0.026 for VAS severity) (figure 1). However, 46% of the purely verbally aggressive incidents had a SOAS-R severity score of ≥ 9 points and were classified as severe incidents, while in contrast 42% of the physically aggressive incidents scored as less severe.

Dependent on their nature and severity, between 43% (verbal aggression) and 67% (SOAS-R-severity score of ≥ 9 points) of the aggressive incidents were followed by a coercive measure, in most cases in form of seclusion or seclusion accompanied by per oral medication or forced injection, while physical restraint was rarely used (table 4).

No aggressive incidents were recorded amongst 1755 (87.0%) of the 2017 patients, while 262 (13.0%) of the patients were responsible for one or more incident, and 6.7% were involved in one or more physical attack. 38 (1.9%) of the patients accounted for 51% of all events, and 58 (2.9%) of the patients accounted for 50% of all attacks (see table 5).

The occurrence of one or more aggressive incident was significantly associated with the age of the patient, admission status of the patient, length of stay and diagnostic group (table 6). A lower risk was found in patients aged over 50 (OR 0.67), in patients with a short length of stay (OR 0.46), in patients with a diagnosis of substance abuse (F1) (OR 0.50), and, in patients with neurotic or personality disorders (F4/F6) (OR 0.56). A higher risk was found in involuntary patients (OR 2.16), in patients with a length of stay of ≥ 17 days (OR 2.72), in patients with a diagnosis of schizophrenia (F2) (OR 2.10) or a diagnosis out of the ICD-10 categories F0, F5, F7, F8, F9, F10 (OR 2.16). No significant association with the risk for aggression was found for gender and affective disorders (ICHD-10 F3) (table 6).

One in four of the incidents and 1/3 of the attacks were registered within the first three days, 37.5% within the first 7 days of the hospitalisations and about half of the incidents within the first 14 days (table 7).

Shift reports of 191 patients covering 573 treatment days were checked by C.A. and I.N. for descriptions of aggressive incidents and then compared with the SOAS-forms from the respective time-periods and patients. We found 11 incidents not registered on the SOAS form. Two of these incidents included physical aggression (in one event, during a dispute between patients one of these threw a cup against his counterpart; in the other event a patient hit a nurse with a bedpan). Nine of the unreported incidents included only verbal aggression, but only in 2 of theses cases the verbal aggression was explicitly described as threatening and was therefore clearly covered by the definition of aggression included in the SOAS-form. According to our estimation of the severity of these incidents, the severity scores were between 1 and 5 points and none of them would have reached a severity score of 9 ore more. Therefore, calculated for the 41560 treatment days in our study, we would have about 280 unreported incidents less severe physical aggression, but no unreported incidents of a severe nature scoring 9 or more points. Considering our total of 681 reported incidents, this would equal a 30% rate of underreporting of "mild" incidents.

The inclusion of 24 wards from 12 hospitals and more than 40'000 treatment days makes our study to one of the largest studies addressing aggression specifically in acute wards. Given the close confidence intervals we consider our results as robust and less subject to local variations in single wards or hospitals. To our knowledge this study is the first to report data on a sample of wards representative for a larger geographic area. The strengths of our study include its prospective nature, the use of a standardized and widely used instrument, enabling us to report different incident rates and rates for different types of aggressive incidents, including confidence intervals which are rarely reported. This is one of the few studies including a systematic check of underreporting enabling us to estimate its extent. However, limitations of this study include the observed underreporting of less severe incidents and of patient-to-patient aggression, and the relatively short observation period of 3 month per ward. The participating wards are a convenience sample and therefore we cannot fully exclude the presence of a sampling bias. However, the sample included one out of four acute wards within the study area. Control for differences in the characteristics of participating and non-participating wards showed some differences. Despite of the inclusion criteria for acute admission wards, a limitation related to studies on acute wards is a possible remaining heterogeneity of the wards due to differences in service organisation (e.g. number of beds per catchment area), hospital organisation (e.g. degree of specialisation of wards) and policy (e.g. referrals from admission to other wards). This hampers the generalisability beyond the wards under study. Another limitation is the lack of more detailed information on socio-demographic, clinical and behavioural characteristics of the patients (e.g. data on the severity of psychopathology and illness). The observation periods of the study wards covered all seasons. However, they were unequally distributed and we had more months of observation in autumn and winter than in spring and summer. Thus we cannot exclude a possible seasonal bias.