Anatomy of a Transformation: A Systematic Effort to Reduce Mechanical Restraints at a State Psychiatric Hospital

The study was conducted at a 398-bed state psychiatric hospital in North Carolina that has five clinical service units: acute adult, community transition, geriatric, child and adolescent, and forensics. This study focused on the adult population (18–64 years) served on the 140-bed acute adult unit (AAU) and the 76-bed community transition unit (CTU), a longer-term rehabilitation unit. The primary populations served on the units are individuals with schizophrenia-spectrum disorders, mood disorders, and substance use disorders. Participants consisted of all persons admitted to the units during the study period (September 1, 2009, to July 31, 2012). The demographic composition of participants remained similar throughout the study. [Tables summarizing data on participant characteristics by unit and study phase are available in an online data supplement to this report.]

We implemented two main strategies to reduce mechanical restraint. The first (phase I) consisted of training all staff in deescalation techniques and forming a response team to assist in crisis situations. The second (phase II) introduced a formal policy change that required additional upper-management approval for the intervention. These strategies were implemented during the study but have become part of the culture and the hospital’s standard operating procedure and have continued beyond the study.

During phase I, we implemented the Crisis Prevention Institute’s (CPI) Nonviolent Crisis Intervention (NVCI) training program (4), which taught deescalation techniques and prevention and management of aggressive behavior. Trainees learned to identify signs of escalating behavior and strategies for avoiding power struggles and for setting limits. They also practiced models to use when confronted with anxious, hostile, or violent behavior, including self-defense maneuvers. The program emphasized that physical intervention is used only as a last resort when there is imminent danger to the individual or others.

NVCI was implemented through a train-the-trainer model. Selected hospital employees were trained by CPI to become instructors, and when trainers were in place, all employees were required to successfully complete 16 hours of initial training and annual recertification thereafter.

To maintain fidelity to the model, a specialized response team of individuals with advanced NVCI training was developed to assist in crisis situations. During standard NVCI training, instructors approached staff members who excelled in the techniques with an offer to participate in the advanced course and for a position on the response team. The added duty required ongoing training, rotating participation on a committee, and commitment of availability to assist in crisis situations.

When a crisis situation occurred, response team members safely rushed to the location indicated by an announcement on the overhead public address system. The unit staff members contained the situation until the response team arrived. This response sequence was patterned after a Code Blue medical response wherein the immediate staff contained the situation until the more fully trained team arrived. The first team member to arrive was considered the lead and was responsible for the success of the deescalation process. He or she was empowered to decide whether a restrictive intervention was indicated. However, the team’s mottos were “de-escalation never stops” and “the best restraint is no restraint at all.” The NVCI techniques and response team were implemented hospitalwide on September 1, 2010.

To ensure that the deescalation principles were maintained and enhanced, we followed a continuous quality improvement model. First, the response team debriefed after each situation to discuss successful aspects and areas for improvement and met quarterly to identify trends. A committee of NVCI instructors and team members also met monthly to ensure that the values, principles, and techniques were effective and preserved. Third, a core team, designated by the chief nursing officer, met with the response team members several times a week and studied the documentation for each response.

After noting success with NVCI and the response team, hospital administrators implemented a second intervention. NVCI and the response team continued to function as described, and the hospital introduced a policy that prohibited routine use of mechanical restraint by requiring staff to obtain permission from the chief medical officer or deputy chief medical officer before using the intervention (phase II). This decision was based on the desire to provide trauma-informed care and the belief that staff had the therapeutic tools needed to deescalate crises without using mechanical restraints. The policy was implemented on CTU on August 8, 2011, and on AAU on September 6, 2011.

In addition to the two main strategies, other factors were present during both phases that likely facilitated our efforts. These factors included leadership, open communication with staff and consumers, and the quality-monitoring procedures for the response team described above. In terms of leadership, the hospital had major buy-in from upper management. The hospital chief executive officer (CEO) questioned the need for mechanical restraint and advocated for the policy to discontinue the practice. Notably, buy-in and commitment extended beyond the CEO to other key staff members. The chief medical officer strongly supported the initiative and was the prime champion for implementing the response team, and the chief nursing officer and director of quality management served as response team members from the beginning of the initiative to show support.

Open communication was also provided to staff and consumers. Regular feedback was provided to staff in the form of e-mails, meeting announcements, and posters displayed throughout the hospital. These communications detailed the hospital’s philosophy and policies regarding restrictive interventions and emphasized the reduction effort. The number of days without mechanical restraint on each unit was celebrated. Consumers also received communication via debriefing within 24 hours after a restrictive intervention. During the debriefing, consumers and staff considered the factors that contributed to the incident and ways to prevent further events.

To examine the effects of the initiative, data were collected between September 1, 2009, and July 31, 2012. Baseline data were collected for one year (September 1, 2009, through August 31, 2010). In phase I of the study, only the NVCI and the response team were implemented (AAU, September 1, 2010, through August 31, 2011; CTU, September 1, 2010, through July 31, 2011). Phase II occurred after the addition of the formal policy change (AAU, September 1, 2011, through July 31, 2012; CTU, August 1, 2011, through July 31, 2012).

Information was compiled from three sources: the Department of Mental Health Enterprise Accounts Receivable Tracking System, the hospital’s administrative database for restrictive intervention use, and the hospital’s pharmacy database. To take into account the fluctuating daily census over the three study years, daily incidence rates were calculated for all variables in the model. These rates were defined as the total number of events (for example, number of persons placed in mechanical restraints) divided by the number of persons at risk for an event (that is, the total number of persons on the unit that day). Mechanical restraint was the primary dependent variable. Four other factors that may influence rates of mechanical restraint were included in the model: seclusion rates, manual hold rates, PRN medications administered for agitation, and assault rates (total number of assaults on patient, staff, and property). The rates for seclusion, manual hold, and PRN use were identified as possible replacement behaviors for mechanical restraint use and were included in the model to determine whether any changes in mechanical restraint use were significant after considering these other variables. The assault rate was included in the model on the basis of hospital data suggesting that assaults tended to precede the use of mechanical restraint.

The numbers of injuries to staff and consumers as a result of assault or containment procedures were also examined to determine whether the number of injuries was affected by the reduction in mechanical restraint use. Because all variables were routinely collected and deidentified, the need for institutional review board approval was waived by the University of North Carolina review board.

Multivariate analysis of variance (MANOVA) was conducted for each unit to evaluate the effect of the study phase on incidence rates of mechanical restraint. Because the Levene’s statistic indicated unequal variances for the variables, we used Tamhane’s T2, a conservative pairwise comparison based on a t test, for all post hoc comparisons. Statistical analyses were performed using SPSS, version 21 (5).

A one-way MANOVA revealed significant multivariate main effects for study phase on both the acute (AAU) and rehabilitation (CTU) units. Given the significance of the overall tests, the univariate main effects for mechanical restraints were examined and were significant on both units (Table 1). The findings indicated that after implementing NVCI and the response team (phase I), the number of mechanical restraint incidents was significantly reduced on both service units (p<.001; AAU, N=2,910 consumers across all study phases; CTU, N=334 consumers). After the addition of the policy change (phase II), a significant decrease was noted on AAU (p<.001); no decrease was noted on CTU because mechanical restraints had already been eliminated. Mechanical restraint use decreased by 98% on AAU and by 100% on CTU. At the end of the study, CTU had not used the intervention in 559 days.

AAU also experienced a decrease in rates of seclusion and of PRN medication use by the end of the study, whereas CTU experienced an increase in both variables. Neither unit experienced a significant change in use of manual holds. For both units, the assault rate was greater at the start of the study period than at the end (Table 1). The number of injuries to staff and consumers as a result of assault or containment procedures was also reduced on both units. [A table in the online data supplement presents information on the number of episodes for each variable by unit and phase.]

Although use of restrictive interventions has a long history in inpatient psychiatric settings, many now recognize the importance of reducing these practices to provide trauma-informed care. Although some call for a complete elimination of both seclusion and mechanical restraint (6), others state that the goal for acute populations should be minimal use because complete elimination may increase the risk of injury to staff and consumers and lead to overuse of medications and premature discharge to prevent high-risk individuals from accruing hospital days (2,7,8). The key is to develop a restraint reduction plan that provides recovery-oriented, trauma-informed care while also minimizing these risks. As others have found (3,6,9,10), we learned that committed leadership was essential for developing and implementing such a plan. Recovery-oriented, strong leaders were champions for the initiative.

Our approach incorporated elements previously identified as being integral for successful restraint reduction programs, including strong support from leadership, formal changes to policy and procedures, staff training, debriefing of consumers, and regular feedback to staff (6), as well as a previous change to a recovery-oriented culture, data-driven monitoring of the initiative, and a focus on continuous quality improvement (3). Moreover, monitoring the performance of the response team and requiring approval for use of mechanical restraint provided a level of accountability for staff actions and encouraged staff to follow the deescalation principles. After implementing these strategies, AAU reduced use of mechanical restraints by 98%, and CTU completely eliminated use of the intervention.

Of note, although we did not focus on seclusion, our efforts had a positive impact on reducing this practice on AAU. Although the rates of seclusion increased on CTU during the study period, the reduction in use of mechanical restraint remained significant after the analysis controlled for the increased rates of seclusion. It is important to note that the reduction in mechanical restraint use on both units occurred without increased rates of assaults or injuries to consumers or staff.

It is unclear why PRN use of medications significantly increased on CTU but not on AAU. Consumers on the units or the prescribing habits of psychiatrists on the units may differ. Further investigation of factors that influence prescribing and administering PRN medications seems warranted. The fact that both units were able to decrease mechanical restraint use, however, suggests that increased PRN medication use is not required to achieve this outcome.

The results of this study are encouraging, but the study had limitations. Around the start of phase I, the hospital merged with another psychiatric facility that was closing, and consumers and staff from the other hospital were transferred to our location. Although we controlled for the increased census by examining incidence rates for the variables, the merger or other historical events over the three-year study may have had an impact on the variables. Also, the study had no control group. It occurred at one state psychiatric hospital in North Carolina, and the results may not generalize to all settings.

The success of this initiative demonstrated that reduction and even elimination of mechanical restraint can be accomplished in a state psychiatric hospital on both acute and rehabilitation units without increasing assaults and injuries to consumers or staff. In addition, our efforts resulted in reduced use of seclusion on the acute unit. The findings suggest that a reduction in restrictive interventions is possible through staff training in deescalation techniques, use of a response team for crisis situations, and policy changes. Other important factors included supportive leadership, feedback to staff and consumers, and quality-monitoring procedures. These strategies have since become part of the hospital’s culture and standard operating procedure, which is important for lasting change. Since the end of the study, AAU has joined CTU in being free of mechanical restraint use for over 12 months.

The authors report no competing interests.