This manuscript was written in accordance with Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines (Chan et al 2013).
The DEMEANOR study is a single-center, pragmatic, cluster, crossover trial, testing the superiority of the BIT program to usual care, beginning 1 March 2019 on one medical surgical unit and one mixed cardiac care clinical unit at Vanderbilt University Medical Center in Nashville, TN, USA. The behavioral intervention team is deployed to each unit in alternating months, as shown in Figure 1. Patients presenting to the unit during a month when the team is present contribute to the intervention arm, while those presenting to a unit when the team is not present contribute to the control arm.
The two study units are a 27-bed mixed cardiac medical surgical unit and a 22-bed cardiac stepdown unit.
All adult (aged ≥ 18 years) patients admitted to either of the two units during the study period are eligible for the study. To prevent contamination between study arms, patients must both be admitted to and discharged from the unit during the study month; patients present on a unit at the crossover will not be included in the analysis. In addition, all nursing staff working in these units during the study are included.
Patients presenting to and discharged from participating units are enrolled automatically. As a cluster, crossover trial, randomization was limited to which unit the intervention would be deployed to first. Unit nursing staff were invited to participate in online surveys by e-mail, with reminders through staff meetings; their participation is voluntary.
The integration of the behavioral intervention team into the unit is not itself a research intervention. Institutional leadership planned to deploy one team as a demonstration project, and to scale it if successful. This provided an opportunity for rigorous evaluation of the team’s effectiveness. The behavioral intervention team, a multidisciplinary team comprised of a psychiatric-mental health advanced practice nurse and a clinical social worker, with psychiatrist consultation as needed, is a modified version of an established model at Yale New Haven Hospital (Sledge et al 2016). The team proactively screens patients upon admission to the unit utilizing a protocol which includes a comprehensive chart review, and if indicated, a brief interview, seeking to identify those patients who possess risk factors indicative of either a potential psychological barrier to their own clinical progress or a potential risk for exhibiting disruptive, aggressive or self-injurious behavior during their hospitalization. Once identified, the team provides interventions aimed at mitigating these risks, through a variety of patient-specific interventions including:
- Psychiatric consultation and recommendations for symptom management
- Behavioral plans of care for nurse/patient interaction
- Psychosocial support and brief psychotherapeutic intervention
- Curbside consultation for any member of the patient’s healthcare team
- Patient advocacy and care coordination
- Psychiatric-specific disposition support, including both inpatient and outpatient psychiatric services
The team also provides education and support to the patient care teams (nurses, physicians and others), and assists non-psychiatric staff in the management of patients who require behavioral health care.
All patients on a unit when the behavioral intervention team is present will be eligible for behavioral intervention team services, including those who are already admitted when the team crosses into the unit. During a month when the behavioral intervention team is not present on a unit, the care and management of patients is not supported with proactive screening and management. Unit staff have access to all psychiatric or behavioral care routinely available in usual care. To ensure continuity of care, patients who continue to require psychiatric support at the time the behavioral intervention team rotates off the unit will continue to receive care equivalent to the usual care condition. It is anticipated that monthly rotations will occur for 10 to 12 months.
All unit staff will be exposed to the intervention throughout the study, although after the first month of the study only those staff on the unit initially randomized to the intervention will have been exposed. This provides an opportunity for a comparison between exposed and unexposed staff. We note that patients present on the unit during the time of cross-over are exposed both to the intervention and the control condition. To avoid cross-over effects and eliminate logistical challenges of changing any behavioral care provider team during the admission, patients are excluded if they are present on a unit during cross-over. This does lead to some possibility of contamination since the behavioral intervention team may be present in a unit during a control team. However, the behavioral intervention team comprises just two providers who were provided with strict instructions to avoid such contamination and who were provided with training on how to defer questions or requests that would have resulted in contamination.
Data are being collected for patients and for unit staff. Patient information is documented as a component of usual clinical care, and those needed for this study will be extracted from the electronic health record (EHR). Data include demographics, documentation of patient disruptive behavior and associated medication administration, and patient length of stay. Disruptive behavior was not previously specifically defined as part of routine clinical care. Documentation in the clinical chart was variable. As part of the study, new documentation fields were added to the electronic medical record to enable description (and subsequent extraction) of specific patient behaviors. Disruptive patient behavior will be documented including physical behaviors (hit, kick, slapped, shoved, spit at, thrown objects, grabbed, bitten, or attacked), verbal behaviors, (threatening, bullying, harassing, name calling, blaming, insulting, yelling, cursing, intimidation), and any clinical interventions that are implemented to control any violent behavior.
Data will be extracted from the EHR by an automated query. The query will be validated by manual cross-check of a subset of cases. Data will be identified using an encounter number, allowing for merging data for individual patients from different forces. The data files generated by manual query will be stored on secure servers and will be merged into a single dataset for analysis by the statistical team.
Anonymous surveys will be used to measure staff perceptions of their ability to manage disruptive patient behavior, and their experience with the behavioral intervention team. The behavioral health survey previously used at the organization is a 10-item instrument that assesses staff perceptions of workplace violence and safety (Table 1). The ten-item scale was adapted from prior work by Loucks et al. (2010) and Rutledge et al (2012) who have validated the Behavioral Health Care Competency Survey. The survey was evaluated for content validity for the current purpose. For this study, we have adapted these instruments to be administered to nursing staff on the two participating units at three different time points: prior to the study, following the first one-month intervention period, and after the study period has been completed (Figure 1). The surveys will be conducted electronically using Research Electronic Data Capture (REDCap) (Harris et al 2009). REDCap allows for surveys completed by the same individuals to be linked while maintaining the anonymity of respondents by preventing access to the linkage tables held securely within the REDCap application.
This study has two primary outcomes: i) any documented interventions to manage disruptive, threatening or acting out behavior, and ii) staff self-reported comfort with and ability to manage patients exhibiting disruptive, threatening or acting out behavior. Documented evidence of intervention is defined as:
- violence control
- injurious behavior
- administration, including as needed administration of the following medications for behavior management: Quetiapine (Seroquel), Alprazolam (Xanax), Clonazepam (Klonopin, Haloperidol (Haldol), Lorazepam (Ativan), Olanzapine (Zyprexa), Risperidone (Risperdal), and Ziprasidone (Geodon).
Secondary outcomes include the occurrence of each intervention separately, sitter use, use of physical or chemical restraints, patient length of stay, and unit nursing staff turnover (Table 2).
Approximately 17.5% of patients admitted to the two participating units required one or more of the behavioral interventions in the year prior to the study. To determine the number of months of data collection necessary to detect a reduction of 5% in the use of behavioral interventions between the intervention and control arm, we used the method described by Arnup et al. (2017). Based on patient length of stay and admit rates, we estimated between 89 and 105 patients would be eligible for inclusion each month (patients present on a unit at crossover will not be included in the analysis). Therefore, a cluster size of 90 was used. If the within period correlation is 0.1 and the within cluster within period correlation is also about 0.1, we would achieve 80% power with 10 months of data collection (total sample size of 1790 patients exposed to either the control or intervention arm). Since these estimates are based on assumptions and it is possible that either a shorter or longer duration is needed, we have specified an interim analysis at 6 months to confirm the overall event rate, and to estimate the observed correlations. No comparisons between groups will occur at this time; accrued data will be used only for sample size re-estimation.
Data and safety monitoring and interim analysis
The study involves the collection of data to compare patients who are or who are not exposed to a clinical care practice: the behavioral intervention team. As such, there is no research intervention. The PI and Co-PIs, in conjunction with the clinical nursing unit leadership, clinical nurse leaders and clinical nurse educators, will monitor for any potential impact of the study that may cause untoward impact on patients or staff on an ongoing basis throughout the duration of the study. If any untoward impact is observed, it will be reported to the study investigators and the IRB.
At six months, the data will be reviewed to determine how many additional months of data collection will be required to have sufficient power to detect the 5% reduction in behavioral outcomes that is being designated as minimally important. Based on power calculations, six months is the minimum period of time for which the study can run to find this difference. At six months, there will be sufficient information to more precisely estimate the magnitude of the within cluster and within cluster within period correlation to update the sample size estimate. Formal stopping rules for safety, efficacy or futility have not been designated.
Statistical analysis principles
There are two main analyses. The first will compare outcomes between patients exposed to the behavioral intervention team and patients not exposed to the behavioral intervention team. The outcomes are quantitative variables that are either binary or ordinal in nature. Comparisons will use a logistic regression or proportional odds models, adjusted for covariates. The models will include cluster (unit) as a random effect, and will also include period including the following patient characteristics: age, sex, race, comorbidities, reason for admission, and psychiatric diagnoses.
The second analysis will compare staff perceptions between those exposed to the behavioral intervention and those not exposed to the behavioral intervention during the first month, and will also compare perceptions between before exposure to the behavioral intervention and after the study concludes. We will compare categorical variables using chi-square tests, and we will compare continuous variables using paired or independent samples t-tests as appropriate. For continuous data that are non-normally distributed, and for ordinal variables, we will use the Mann-Whitney U-test or the Wilcoxon test as appropriate.
While our initial approach is to a simple comparison, this ignores the repeated measurements. Since staff on the unit can leave and new staff can join the unit, we will use a mixed-effects model to model differences between before, at one month, and on completion of the study. Respondents will be included as a random effect. Linear, logistic or proportional odds model will be used as appropriate for the outcome. Finally, we may model the survey outcomes using similar regression techniques to explore factors associated with changes in staff perceptions.
Treatment of outliers
Our evaluation of whether an individual patient experiences an event or not reduces the impact of outliers on the primary outcome. For survey results, the results are constrained to the available options, again limiting outliers. Therefore, any statistical modeling will not evaluate the impact of outliers. There is the possibility that there are outliers for variables such as length of stay, or for counts of behavioral problems. These are expected and will be reported. These outcomes will generally be compared between groups or modeled using non-parametric approaches, which are robust to the presence of such outliers.
Presentation of the results
After completion of enrollment and data analysis, the results of the trial will be communicated to the public through manuscript publication and submission of the results for the ClinicalTrials.gov database. Submission for publication will include public access to the full study protocol and statistical code. Authorship will be based on the International Committee of Medical Journal Editors guidelines (2018), and professional writers will not be used.
The flow of patients through the study will be presented in a flow diagram (Figure 1). Baseline characteristics will be presented by treatment group, as shown in Table 2). The percent of patient injurious behaviors reported will be determined by extracting reports of biting, kicking, throwing, etc. from the EHR. The percent of violence control measures used will be determined by extracting medications, restraints and sitters ordered from the EHR.
A composite of all documented indications of disruptive patient behavior including violence control nursing interventions, as needed (PRN) medication administration for behavior management, nursing problem of violence risk, or any injurious behavior will be extracted from the EHR. Each component of the primary outcome will also be reported separately, along with documented physical restraints for disruptive patient behavior.
The percent change in nurse comfort and confidence in their ability to manage patients exhibiting disruptive, threatening or acting out behavior, perceptions of workplace violence/safety, as well as perceived confidence, comfort and most helpful training/support received will be determined by comparing pre-, one month and post-implementation surveys.
The rate of unit nursing staff retention will be determined by extracting rates from internal human resources records. Patient length of stay will be determined by extracting data from the EHR.