To compare the effectiveness of strategies to prevent and de-escalate aggressive behaviors in psychiatric patients in acute care settings, including interventions aimed specifically at reducing use of seclusion and restraint.
We searched MEDLINE®, Embase®, the Cochrane Library, Academic Search Premier, PsycINFO, and CINAHL from January 1, 1991, through February 3, 2016. We manually searched reference lists of pertinent reviews, included trials, and background articles to identify relevant citations that our searches might have missed. Eligible studies included randomized controlled trials (RCTs), cluster randomized trials (CRTs), and observational and noncontrolled studies with sample sizes greater than 100. Eligible studies were limited to acute care settings and adult patients with psychiatric disorders or severe psychiatric symptomatology (excluding dementia); they had to report on aggression or seclusion and restraint outcomes.
Two investigators independently selected, extracted data from, and rated risk of bias of studies. Risk of bias and strength of evidence (SOE) were assessed only for controlled studies. Twenty-nine primary studies (from 31 articles) met inclusion criteria. Of these, 11 were controlled trials that provided eligible data for SOE grades. Only 4 of these trials took place in the United States. We grouped studies as follows: (1) staff training interventions, (2) risk assessment interventions, (3) multimodal interventions, (4) environmental interventions (including group psychotherapeutic options), and (5) medication protocols versus other medication protocols or alternative strategies. We organized results by three key questions; these covered benefits, harms, and potential modifying characteristics of these strategies.
Evidence was limited for benefits and, especially, for harms; information about modifying characteristics was completely absent. No key questions had data supporting SOE grades better than low, indicating limited confidence that the estimate of effect lies close to the true effect for these outcomes. The available evidence comprised primarily pre/post studies whose inherent high risk of bias precludes drawing inferences of causality. Of the 11 trials eligible for SOE assessment, all but 1 had medium (or high) risk of bias. Risk assessment had low SOE for decreasing subsequent aggression and reducing use of seclusion and restraint, but only when applied in a preventive manner (e.g., as unit-wide programs). SOE for all other interventions, whether aimed at preventing aggression or de-escalating aggressive behavior, was insufficient.
Given the ethical imperative for treating all patients with dignity, the clinical mandate of finding evidence-based solutions to these mental health challenges, and the legal liability associated with failure to assess and manage violence risk across the treatment continuum, the need for evidence to guide decisionmaking for de-escalating aggressive behavior is critical. The available evidence about relevant strategies is very limited. Only risk assessment decreased subsequent aggression or reduced use of seclusion and restraint (low SOE). Evidence for de-escalating aggressive behavior is even more limited. More research is needed to guide clinicians, administrators, and policymakers on how to best prevent and de-escalate aggressive behavior in acute care settings.