Patients afferent to GHPWs from March 25^th 2002 to July 26^th 2002 were eligible for the study. Fifteen sites took part in the study. Patients previously enrolled in this study and newly admitted to GHPWs were excluded from participation; however, any new admission was recorded in the case report form.

Descriptive epidemiology included the analysis of diagnoses distribution according to DSM-IV and ICD-9, and the evaluation of social and demographic profile of patient population, the reason of hospitalisation and the interventional procedures. Clinical epidemiology was based on the assessment of prevalence of aggressive episodes at admission and their incidence in the observational period; the evolution of the acute psychotic episode (diagnosis, treatment and outcome) was also evaluated. The outcome of the acute episode was evaluated in patients with one of the following group of diagnoses: schizophrenia, depression, nevrotic disturbance, substance abuse, psychorganic psychoses, mania, undifferentiated, antisocial and non-antisocial personality disorder.

The study design and procedures, including time of assessment, are summarized in Figure 1. The maximal observational length was 30 days; daily recording of interventions and outcome was performed in the first 7 days of hospitalisation. Visits at Psychiatric Wards were scheduled at study entry (day 1, admission), at follow-up (day 7) and at discharge (final visit). On day 30, observation was discontinued anyway and assessments of final visit were performed. In patients discharged prior to or at day 7, forms for final visit were to be completed, without any follow-up observation. A form for the next 5 hospitalisations following that of the present study was also to be completed.

The following psychometric scales were used for assessment: BPRS (Brief Psychiatric Rating Scale), MOAS (Modified Overt Aggression Scale), and NOSIE-30 (Nurses' Observation Scale for Inpatient Evaluation).

BPRS is a validated and widely used psychometric scale. An Italian expanded 24-item version (BPRS version 4.0) was used in this study 9. Severity of a total of 24 symptoms, grouped in 6 different domains, was evaluated using a 7-point rating scale ranging from 'not present' to 'extremely severe' to obtain an overall total score. High levels of inter-rate reliability between experienced (i.e. psychiatrists and psychologists) and inexperienced operators (i.e. medical and psychosocial rehabilitation students) were shown in previous trials 10. BPRS version 4.0 was administered at admission, at day 7 and at discharge (or day 30) and forms were completed by physicians following a patient's interview.

The MOAS scale 11 records the forms of aggression and their severity; it is constituted by 4 subscales based on increasing severity: verbal aggression, aggression towards properties, self-aggression, and physical aggression towards people. Each subscale includes 5 items scored 0-4; the total score is obtained by multiplying scores of each subscale by their specific 'weight' (1, 2, 3 and 4, respectively), then adding the 4 obtained values. Subjects with aggressive behaviours are defined as those having a total score > 0 in the observational period. MOAS was completed by non-medical healthcare personnel to assess outcome or onset of aggressive episodes; it was administered at admission, at day 2, 3, 4 and 7, and at discharge (or day 30).

The NOSIE-30 12 was used to assess frequency of 30 behaviours in hospitalised patients, ranging from 'never' to 'always'. The 30 items are divided in 7 different domains: social competence, social interest, personal neatness, irritability, manifest psychosis, retardation and depression. The first 3 domains reflect positive behavioural dimensions (Total Positive Factors), and the other 4 are indicators of negative behaviours (Total Negative Factors). The Total Patient Asset score was obtained by the sum of the positive factors minus the sum of negative factors and adding 150 as a normalisation factor.

One Local Study Coordinator (LSC) was identified in each participating site. LSC was responsible for study conduction and study material distribution; he/she also identified and properly trained the Clinical Investigators (CI) and Raters, and ensured their correct application of study procedures. CIs were responsible for patients' selection and enrolment, BPRS administration according to protocol, completion of case report forms, and contact with Data Management centre (including quality controls). Raters were charged of MOAS and NOSIE-30 administration and correctness.

Data Management, quality control and Statistics were performed by Runtimes srl, Modena (Italy). Results of parametric variables were presented as means ± standard deviation and range, while results of categorical variables were presented as number and proportions. Evaluable patients were those completing both the admission and the discharge visits.

The general characteristics of patients' populations are summarised in Table 1. A total of 864 patients were enrolled in 15 GHPWs and 728 of them (320 males and 408 females) resulted to be evaluable (i.e. had both admission and discharge visits). The mean age (± standard deviation) was 43.6 ± 14.8 years (range 16-99) in the total population, and was slightly higher in females than in males. Most of patients were included in the age ranges of 26-34 (164 patients), 35-44 (177 patients), 45-54 (131 patients) and 55-64 (108 patients) years. With regards to life habits, most of patients were non-alcohol users (402, 55.2%) and non-drug users (602, 82.7%), while more smokers (385, 52.9%) than non-smokers (291, 40.0%) took part in the study.

The main reason of admission was a severe psychotic episode with (139 patients, 19.1%) or without (347, 47.7%) aggressive behaviour; less frequent reasons included moderate psychoses with unavailability of any caregiver (108, 14.8%) and Axis I disorders with alcohol abuse (56, 7.7%).

The most frequent patients' referrals were the Hospital emergency department (245, 33.7%), a mental-care centre (119, 16.3%) and self-referral (103, 14.1%); 429 patients (58.9%) had a known diagnosis at admission. The mean number of hospitalisations in the previous 12 months was 1.12 ± 2.51.

The mean length of hospitalisation was 12.0 ± 10.2 days (range 1-92); 24.7% of patients stayed in GHPWs for more than 15 days, 17.9% for 11 to 15 days and 16.9% for 8 to 10 days, while lower rates of patients had a shorter stay.

Primary diagnoses at admission and discharge are presented in Figure 2. The most frequent groups of diagnoses (according to ICD-9) at admission were schizophrenia (199 patients, 42.8%), depression (60 patients, 12.9%) and undifferentiated personality disorder (47, 10.1%). Less frequent diagnoses included mania (40, 8.6%), non-antisocial personality disorder (33, 7.1%), nevrotic disturbance (26, 5.6%), psychorganic psychoses (19, 4.1%), substance abuse (18, 3.9%) and antisocial personality disorder (9, 1.9%). Diagnosis at discharge were schizophrenia (268 patients, 40.1%), depression (98 patients, 14.7%), nevrotic disturbance (62, 9.3%), undifferentiated personality disorder (61, 9.1%), mania (52, 7.8%), non-antisocial personality disorder (40, 6.0%), psychorganic psychoses (29, 4.3%), substance abuse (22, 3.3%) and antisocial personality disorder (15, 2.2%). Diagnosis at admission was not available in 263 patients (36.1%), while 59 patients (8.1%) were not diagnosed at discharge.

A total of 59 patients (8.1%) had at least one further hospitalisation after discharge: the mean number of further admissions was 1.39 ± 0.81 and mean duration was 10.68 ± 8.72 days; main reasons were a severe psychotic episode with (9 patients, 11.0%) or without (40, 48.8%) aggressiveness.

The number of patients with at least one episode of aggressiveness was 44 (6.0% of hospitalised) at admission and progressively declined over time: they were 24 (3.4%) at day 2, 13 (1.9%) at day 3, 9 (1.4%) at day 4, and 8 (1.7%) at day 7.

Results of MOAS are presented in Figure 3. A marked and progressive decrease of mean scores from admission to discharge was observed in total score and in each domain. Total score was 2.53 ± 5.1 at admission, 0.38 ± 2.2 at day 7 and 0.21 ± 1.5 at discharge; changes of single domains (verbal aggression, aggression towards properties, self-aggression, and physical aggression towards people) were consistent with those of total score.

Figure 4 shows results of BPRS version 4.0. Total score was 62.3 ± 21.2 at admission (day 1), 52.6 ± 20.7 at day 7 and 44.7 ± 17.3 at discharge. A progressive decrease over time of values recorded at study entry was also observed in each domain (anxiety-depression, thought disorders, isolation-motor retardation, hostility-suspiciousness, hyper-reactivity, and mania) as well as in each of the 24 single items. Improvements at discharge were observed in each group of diagnosis (data not shown) and were of similar extent in all investigated domains.

Results of NOSIE-30 recorded at day 4, day 7 and discharge are presented in Table 2. No clinically relevant changes were observed in any of the investigated domains from day 4 to discharge. Among 'positive' domains, a small increase was observed in social interest, compared with a small decline in social competence; among 'negative' domains, all of them (irritability, manifest psychosis, retardation and depression) showed small improvements at discharge. Mean values of total patient's asset also did not change from day 4 to discharge and no changes were also observed grouping patients by diagnosis.

Table 3 shows the number of patients with at least one psychotic drug taken at study entry, during hospitalisation (day 1, day 4 and day 7) and at discharge.

A number of 472 (64.8%) patients were previously taking at least one drug at study entry, and this amount increased from day 1 (90.8%) to day 7 (95.2%); 686 (94.2%) patients received drug therapy during hospitalisation and 676 (92.9%) had at least one psychotropic drug prescribed at discharge. Monotherapy was administered in 117 patients (24.8%) prior to admission, decreased during hospitalisation (13.2% at day 1, 7.2% at day 4 and 8.2% at day 7), and was prescribed in 91 patients (13.5%) at discharge, in favour of a more intensive treatment that included combined therapies with increased frequency. Rates of patients with two psychotic drugs taken at admission and day 7, and prescribed at discharge, were, respectively, 22.0%, 20.6% and 26.6%; the corresponding figures of polytherapy with more than 2 drugs were 53.2%, 57.8% and 59.0%.

As a consequence, the mean number of psychotic drugs taken simultaneously was 2.6 ± 1.3 prior to admission and increased to 2.9 ± 1.2 at day 1, 3.2 ± 1.3 at day 4, and 3.3 ± 1.4 at day 7; the mean number at discharge was 2.9 ± 1.2.

The most frequently psychotic drug classes taken at study entry, and prescribed in the first 7 days of hospitalisation and at discharge, are presented in Table 4. Benzodiazepines were the most frequently used drugs at study entry (60.0% of patients); their administration was intensified during hospitalisation (85.7%) and were prescribed in 69.5% of patients at discharge. Rates of inpatients treated with other drugs also increased: typical antipsychotics were taken at study entry and during hospitalisation in 48.3% and 57.0% of patients, respectively, and were also prescribed at discharge in 49.6%; the corresponding figures were 35.6%, 41.8% and 39.8% for atypical antipsychotic drugs, 40.9%, 48.8% and 43.2% for antidepressants, and 23.9%, 27.1% and 29.1% for mood stabilizers. Less frequently used other drugs included antidotes for drug of abuse.

In the overall population, the most frequently used drugs prior to study entry were haloperidol (25.6% of patients), delorazepam (20.1%) and lorazepam (19.5%); predominant drugs during hospitalisation were delorazepam (35.9%), lorazepam (26.7%) and haloperidol (25.4%), while the most frequently prescribed drug at hospital discharge were haloperidol (26.3%), delorazepam (25.7%) and olanzapine (21.2%). Predominant combined therapies administered in inpatients were benzodiazepines-antidepressants (15.6% of patients), atypical drugs-benzodiazepines (14.9%) and benzodiazepines-antidepressants-atypical drugs (9.5%).