The Effectiveness of User-driven Intervention to Manage Patient Aggression in Mental Health Services

This is a single-blind, two-arm, cluster randomised controlled trial involving 28 psychiatric hospital wards across Finland.

Units are randomised on an equal basis to receive either staff educational programme (designed together with staff members, patients and relatives on the study units) or standard care. Inpatient psychiatric hospital organisations are the unit of randomisation. The cluster design will be used to avoid contamination between individual staff members in each study ward. We will use centralized randomisation at the University of Turku (Department of Mathematics and Statistics). Randomisation will be fully concealed and computer-generated by an independent statistician, who is not involved in the study. Investigators enrolling wards cannot foresee assignment. Outcome assessors and statisticians will be kept blinded to allocation. However, due to the type of intervention, allocation will be unmasked to patients and their relatives, contact persons in each ward, and health care staff delivering patient care on the wards after randomisation; this would reflect real-world care. While the Data Monitoring Committee undertook ongoing safety surveillance, investigators running the preliminary analysis for the Data Monitoring Committee will be masked to data until investigators released the database. Further, the statisticians and the National Register holder who are responsible in Finnish routine data used in this study, will be masked to ward allocation and patient data in each group. Randomisation will be done after collecting the baseline data from each unit.

Baseline data on coercive practices used on the units will be collected by the members of the research team with a help of staff members within the eight weeks, while observational and interview data related to the quality of the services will be collected by the research team (only in the intervention units). On the individual level, all staff members working in the study units, patients, and their relatives are eligible to participate in the study. Staff survey and interviews with staff members, patient groups and relatives will be done on the basis of free will. The staff of the unit/research team members will approach patients and their relatives to explain the purpose of the study and its arrangements orally and in written format. The opportunity to ask questions about the study will be given to them. Those that will be assessed having a lack of capacity to give informed consent and declined to participate will not be interviewed. Staff surveys will be repeated for baseline and follow-up. On the organisational level, the outcomes related to patient service use will be assessed 12 and 24 months after baseline data collection.

Assessing quality components: The strengths, weaknesses, opportunities, and threats related to the new practices will also be discussed and categorised based on SWOT analysis (Strength, Weakness, Opportunity, Threat). Possible barriers and facilitating factors for change on the on each unit will be identified. The quality of the service facilities will be assessed by observations on the units, based on World Health Organization's (2012) QualityRight Tool Kit focusing on five central quality criteria.

We have systematically searched for, but found no directly relevant past work. We assume that if the data will be obtained from local hospital registers by sampling 7 clusters (hospitals) with 265 subjects each in intervention group and 7 clusters with 265 subjects each in control group, we will have 80% power to detect a difference between the group proportions of -0.0400. The control group proportion is 0.1100. The intervention group proportion is assumed to be 0.1100 under the null hypothesis and 0.0700 under the alternative hypothesis. The test statistic used is the two-sided Z test (unpooled) with an overall significance level at 0.05. We may assume based on the hospital registers that sample size for the total population admitted in the study wards in one year will be 3710. However, if we consider a loss of 20% patients in the local care registers, the total number of patients on the randomised wards should be about 4454 patients. Further, if we assume 50% response rate for patient survey out of possible 3710 participants, we will assume that we will obtain 928 filled questionnaires during 6 month survey data collection period. The sample size calculation was adjusted for intra-cluster correlation at 0.005.