Evaluation of Individualized De-escalation Strategies for an Individual With Dangerous Behavior

Abstract

De-escalation strategies, or procedures that result in rapid cessation of dangerous behavior, may be needed when the behavior presents an immediate risk of harm to the individual or others. Most de-escalation recommendations are general and not individualized on the basis of the individual or circumstances. This study evaluates an assessment comparing five de-escalation strategies for an individual who engaged in dangerous behavior. Two strategies resulted in shorter durations of behavior episodes relative to other strategies, suggesting a promising avenue for additional study focused on individualizing de-escalation recommendations.

Individuals with developmental disabilities are at increased likelihood of behavior that puts themselves and others at risk of physical injury, such as aggression or self-injury (hereafter referred to as dangerous behavior). Dangerous behavior may limit access to inclusive, educational, or social settings (Myrbakk & von Tetzchner, 2008) or result in injuries, restraint, or seclusion (Mohr et al., 2003).

Dangerous behavior is best treated with function-based behavioral interventions that identify functional reinforcers maintaining the behavior and re-arrange environmental contingencies, often through differential reinforcement, to promote adaptive behavior and decrease dangerous behavior. Although this approach should be a first-line strategy, there are times when other strategies, such as de-escalation, are needed. De-escalation has multiple definitions but generally refers to responding to early signs of dangerous behavior with consequences that prevent the behavior from escalating (e.g., Couvillon et al., 2010; Metoyer et al., 2020). There are several situations in which de-escalation strategies may be necessary. Behavioral treatment is not always readily available, with clinics that treat severe forms of dangerous behavior being geographically limited with long waits. De-escalation strategies may be necessary while waiting for more robust services. Once in services, treatment development may take weeks or months for some individuals. De-escalation may maintain safety during that time. Last, de-escalation strategies are commonly used in settings without extensive behavioral supports (e.g., hospitals, schools; Couvillon et al., 2010).

There are several ways in which de-escalation may impact behavior mechanistically, depending on the strategy selected and the function of the behavior. De-escalation may involve providing hypothesized functional reinforcers to prevent escalation of unsafe behaviors—in other words, in some cases, de-escalation strategies may entail doing the opposite of what function-based interventions often recommend (Call et al., 2022). Other times, it may prevent the occurrence of dangerous behavior by responding to early signs of escalation and reinforcing approximations of coping or adaptive behavior. Common de-escalation strategies involve the addition or removal of stimuli based on individual preferences and functional reinforcers (Stevenson et al., 2019), such as vocal attention that is calm and validating (e.g., “I see you are upset”), physical attention (e.g., a pat on the back), physical space between the individual and others, access to a calm or quiet space, or providing tangible items, often as a redirection, distraction, or for sensory input (e.g., toys, electronics, weighted blanket, fidget toys). Strategies also include removing sensory stimuli from the environment, such as reducing lights, noise, or social interaction.

Considering the variability in functional reinforcers and preferences, no single strategy is universally effective. For example, physical attention can be the evocative event for the dangerous behavior of some individuals (i.e., social avoidance function); thus, creating space may be an efficacious de-escalation strategy. Other individuals may prefer physical attention and engage in dangerous behavior to access it (i.e., attention function), thus creating space that may further escalate the behavior in the moment. Even when the functional reinforcer is known, it cannot be delivered sometimes, and alternative de-escalation strategies are needed (e.g., a broken tangible item). In these situations, it may be helpful to identify non-function-based de-escalation strategies.

Unfortunately, there is a lack of resources to identify individualized de-escalation strategies, especially when the functional reinforcer cannot be delivered. The extant literature suggests practitioners use “on-the-job” training instead of a formal assessment when selecting de-escalation strategies (Dalton et al., 2023). This practice may not reliably yield effective de-escalation across all individuals, especially in the case of complex learning histories. The present study describes an assessment designed to identify effective de-escalation strategies for a client who displayed dangerous, aggressive behavior in an intensive outpatient program. We aimed to identify a de-escalation strategy that was (a) reliably effective in decreasing the duration of dangerous behavior, (b) individualized to the client, and (c) feasible in natural environments.

Method

Participant and Setting

Ricky was a 12-year-old boy diagnosed with autism spectrum disorder. He was admitted to an intensive outpatient program (IOP) for the treatment of aggression that had previously resulted in injury to others and removal from school. Ricky spoke in complete sentences and followed multi-step instructions. Per his individualized education plan, his intelligence quotient was significantly below average (composite standard score of 51 on the Reynolds Intellectual Assessment Scales-Second Edition; Reynolds & Kamphaus, 2015). In the IOP, behavioral assessment and treatment sessions were conducted five hours a day, five days a week, in a padded session room with a table, chair, and low-preferred items. A minimum of three staff members (i.e., registered behavior technicians) worked with Ricky to ensure safety during behavior crises. Staff wore helmets, chest guards, arm guards, padded gloves, and a balaclava for safety. Despite the use of protective equipment, aggression regularly caused concussions, bruising, and open cuts to staff members. Unstructured observations by the supervising BCBA and his team prior to the de-escalation assessment suggested that physical management (e.g., blocking hits) appeared to prolong or escalate aggression and that staff used various de-escalation strategies (e.g., minimizing attention, offering alternative items) during behavior crises with variable success. Over the course of Ricky’s admission, outcomes from a functional analysis indicated that his aggression was maintained by tangible, escape, and attention contingencies. His primary admission goals included re-admission to public school. Thus, factors related to a school setting were factored into the design of the de-escalation assessment (e.g., presence of other students, restraint and seclusion laws, staffing ratios).

Response Measurement

The primary dependent variable was the duration of dangerous behavior episodes. In the de-escalation assessment, de-escalation strategies were implemented contingent on the onset of a behavior episode. A behavior episode was defined as a single instance of aggression or any of the following happening within 15 s: (a) three instances of self-injury or property destruction, (b) three violent statements with screaming, and/or (c) three instances of violent gestures with screaming. The behavior episode was operationally defined in collaboration with staff members who worked directly with Ricky, who identified self-injury, property destruction, screaming, and violent statements/gestures as hypothesized precursors to aggression. Calm criteria (the end of a behavior episode) were defined as two min without any of the previously listed response topographies (see Table 1). During the de-escalation assessment, observers collected duration data on behavior episodes from the onset of an episode to Ricky meeting calm criteria.

Table 1.

Operational definitions for dangerous behavior

Topography	Operational definition
Aggression	Any attempt or instance of hitting, kicking, biting, pulling hair/clothing, scratching, and/or throwing an item at another person
Self-injurious behavior	Any attempt or instance of hitting himself from ≥ 6″, hitting his head against surfaces from ≥ 4″, and/or scratching himself
Property destruction	Any attempt or instance of breaking items, throwing furniture/items, or hitting/kick surfaces
Screaming	Any instances in which the client emits vocalizations above a conversation level
Violent statements/gestures	Any instance in which the client makes a statement or gestures related the use of guns, weapons, and/or making threats of violence towards another individual (e.g., threatening to hurt someone, punching at someone without making contact)

During each day of Ricky’s admission to the IOP, staff collected frequency data on aggression that occurred both in and outside of de-escalation assessment sessions. Given the highly variable nature of behavior episodes (e.g., one or two episodes occurring each week with otherwise low rates of aggression), frequency data were transformed into a rate measure (responses per minute) and summarized across weeks. Rates of aggression during each week of Ricky’s admission served as an additional way to evaluate the effectiveness of the optimal de-escalation strategy using an ABAC design (A = treatment as usual, B = treatment + de-escalation assessment, C = treatment + optimal de-escalation strategy).

Inter-observer Agreement

An independent observer recorded the duration of behavior episodes for 30% of de-escalation assessment sessions. Proportional duration inter-observer agreement (IOA) was calculated by dividing the smaller duration by the larger duration and multiplying by 100, yielding a mean agreement of 93.76% (range 65.851–100%).

De-escalation Assessment Procedure

The de-escalation assessment was conducted during the treatment evaluation phase of Ricky’s admission. In other words, a treatment plan based on functional analysis outcomes was in place each day Ricky attended the IOP, beginning the second week of his 88-week admission. The de-escalation assessment was conducted between weeks 50 and 54 of his treatment evaluation phase, during which staff implemented his treatment according to his treatment plan and initiated a de-escalation assessment session contingent on the onset of a behavior episode. In other words, upon the start of a behavior episode, the staff member paused typical treatment contingencies and implemented one of the de-escalation strategies (described below). Upon calm criteria, the staff member resumed typical treatment contingencies.

De-escalation strategies included in the assessment were quasi-randomly assigned across behavior episodes using a without-replacement method across four series. For example, if two behavior episodes occurred within the same day, different de-escalation strategies were used during each episode according to a pre-determined, quasi-randomized order. De-escalation strategies were selected based on discussions with Ricky’s caregivers and staff and observations of past behavior episodes. If Ricky did not meet calm criteria within 10 min, staff terminated the de-escalation session and used varied strategies based on clinical judgment to help Ricky achieve calm (this often involved a mixture of condition components or whichever condition appeared to be most effective up to that point). Across the entire day, including all conditions of this assessment, staff implemented physical management procedures outlined in the Marcus Crisis Prevention Program (MCPP; Bernstein et al., 2024) as needed to maintain safety when aggression occurred (e.g., blocking hits and bites while otherwise minimizing attention to aggression). Staff attempted to evade aggression and use the least physically intrusive blocking methods possible.

Attention Differential Reinforcement of Other Behavior (DRO)

The staff member stepped to the other side of the room and stated, “I am here for you when you are calm.” If Ricky initiated social interaction, the staff member reminded him they could talk once he was calm on a variable-time one-minute schedule. These procedures were conceptualized as a DRO in that high-quality, conversational attention was withheld until Ricky met the calm criteria. There were no programmed contingencies for dangerous behavior.

Attention Differential Reinforcement of Alternative Behavior (DRA)

The staff member remained in close proximity to Ricky and responded to all social initiations (e.g., verbal or gestural) by either (a) answering a question, (b) telling him they were sorry he was upset, (c) asking him if he wanted to talk, or (d) providing physical attention (e.g., head scratches, back rubs). There were no programmed contingencies for dangerous behavior.

Noncontingent Access (NCA) to Quiet Space + Attention DRO

The staff member brought a weighted blanket and beanbag into the room, stepped to the other side of the room, and told Ricky, “Your quiet space is available to calm down.” If Ricky initiated social interaction, the staff member reminded him that his quiet space was available to calm down on a variable-time, one-minute schedule. There were no programmed contingencies for dangerous behavior.

NCA to Tangibles + Attention DRO

The staff member moved to the other side of the room and asked if there were any items that Ricky wanted. They also provided choices for items (snacks, pop-it toy, beanbag) and reinforced requests for items when possible. Staff limited verbal attention to only offering tangible and food items. If Ricky initiated social interaction, the staff told Ricky, “We can hang out once you are calm, but you can have your items if you want.”

Prompted Coping Skills

Prior to the assessment, staff informally taught Ricky several coping strategies (e.g., hugging a pillow, deep breathing) using instructions, modeling, prompting, and praise outside of behavior episodes. When implementing the Prompted Coping Skills strategy, the staff member presented coping materials, verbally prompted Ricky to engage in coping skills, and modeled the skills if Ricky did not engage in them. The staff member limited other verbal attention. If Ricky initiated social interaction, the staff told Ricky, “We can hang out once you are calm, but we can do your coping skills for now.” The staff provided praise and encouragement for engaging in coping skills. There were no programmed contingencies for dangerous behavior.

Results

The results of the de-escalation assessment are presented in Fig. 1. De-escalation assessment sessions were terminated (i.e., behavior episode lasted longer than 10 min) for 100% of Attention DRA sessions, 75% of NCA to Tangibles + Attention DRO sessions, and 75% of Prompted Coping Skills sessions. Ricky calmed within 10 min of all Attention DRO and NCA to Quiet Space + Attention DRO sessions. The average duration of behavior episodes was not meaningfully different between the Attention DRO condition (M = 6.52 min) and the NCA to Quiet Space + Attention DRO condition (M = 5.83 min). The Attention DRO strategy was selected in conjunction with stakeholders (Ricky’s caregivers and school staff) as Ricky’s optimal de-escalation strategy. Although the Attention DRO and NCA to Quiet Space + Attention DRO strategies produced similar effects, the latter required additional items in the room (i.e., a weighted blanket and beanbag), which could make it more onerous to implement with integrity in varied settings. Ultimately, the Attention DRO strategy was incorporated into Ricky’s final treatment package, resulting in a reduction in rates of aggression (Fig. 2) during his admission to the IOP program. A brief reversal to treatment as usual produced an immediate increase in the rate of aggression, which again decreased once the Attention DRO was re-introduced. Within this final treatment package that included the Attention DRO de-escalation strategy, low levels of aggression were maintained outside of an individual padded session room (week 59), in a room with peers (week 63), and when staffing was reduced to a 1:1 ratio (occurring week 76, not depicted).

Fig. 1.

Duration of behavior episodes during de-escalation assessment. *Note. Bars represent the average duration of behavior episodes within a condition, whereas data points represent the duration of behavior episodes in a given session

Fig. 2.

Weekly rates of aggression during admission. *Note. The red dotted line denotes an 80% reduction from baseline levels of aggression. Data begins at week 34 as that was the last point at which significant changes to the treatment plan were made besides initiation of the de-escalation assessment

Discussion

Direct assessment may afford practitioners a systematic method to individualize the selection of de-escalation strategies for individuals who engage in dangerous behavior (Stevenson et al., 2019). As a preliminary example of one such assessment, our data suggested that two strategies (Attention DRO, NCA to Quiet Space + Attention DRO) resulted in reliably lower duration behavior episodes than other strategies. Of note, analysis at the session-by-session level shows increasing trends for the Attention DRO and NCA to Quiet Space + Attention DRO toward the end of the assessment. To align with the goal of a brief assessment to inform the selection of de-escalation strategies, the assessment was limited to four series, and aggregated data were analyzed. This aligns with standard practices for other types of assessments (e.g., preference and demand assessments; Call et al., 2016). However, future evaluations should consider more systematically determining the number of series needed to maximize confidence in outcomes, including accounting for trends across series.

Compared to treatment as usual, both the assessment itself and the final de-escalation strategy selected produced lower rates of aggression. The reduction during the assessment may be attributed to the most efficacious de-escalation strategies being implemented on at least some occasions, or it could be that all strategies, when systematically implemented at early signs of a behavior episode, contributed to reductions in severe episodes (e.g., Fritz et al., 2013). Interestingly, the two optimal strategies were similar in that both had an attention DRO component. It was reported that caregivers would often deliver attention in the form of soothing during behavior episodes, which may have prolonged periods of dangerous behavior. These findings highlight the importance of systematic assessment, which may counter intuitive judgment.

Although promising, this demonstration is a first step that requires additional research to confirm utility and efficacy. This evaluation is limited to a single individual, and procedural fidelity data are unavailable. We also do not have IOA data for aggression that occurred outside of assessment sessions. Although we involved caregivers in identifying de-escalation strategies, future work in this area should, if possible, involve the individual in developing de-escalation strategies and formally assessing social validity. We also only examined rates of aggression as the outcome of interest within the reversal design. Future research may examine the effect of de-escalation procedures on other presenting topographies, including precursors.

Additionally, it will be important that future researchers evaluate the long-term effectiveness and maintenance of treatment packages that employ de-escalation strategies in conjunction with function-based treatment components—as well as the long-term feasibility and social validity of such packages. Evaluations of this assessment to guide de-escalation strategy selection in the absence of a treatment plan are also needed to determine if this approach is useful when individuals are waiting for or unable to access behavior services. Despite its limitations, the present evaluation provides preliminary evidence that the de-escalation assessment described herein may yield helpful information to clinicians seeking to identify effective, individualized de-escalation strategies and lays the groundwork for additional research on this approach.

Author Contributions

SB – conceptualization, methodology, visualization, writing, review and editing. JH – conceptualization, writing, review and editing. MS – writing, review and editing.

Funding

There was no funding for this project.

Data Availability

Data may be available upon reasonable request.

Declarations

Ethical Approval

This study was approved by the first author’s second affiliation as a retrospective chart review. Waiver of consent was obtained as the study posed no to minimal risk to participants and used data that was previously collected as part of the participants’ admission to the clinical program – no additional data or tasks were completed by participants for the purpose of this project. Clinical consent was obtained by caregivers at the time of the admission. All research procedures followed best practices for human subject research as outlined the Declaration of Helsinki on Medical Research.

Competing interests

All authors are clinical providers that deliver behavioral outpatient services to autistic youth in the service setting described within this study. Each actively provides services and supervises clinicians within that clinic.