Teaching Safety Skills to Children: A Discussion of Critical Features and Practice Recommendations

Abstract

Children may encounter safety threats from the physical environment (e.g., firearms or poisonous substances) or from the behaviors of others (e.g., abduction or sexual abuse lures). Such encounters may result in injury or death if children do not learn skills to respond safely. Research over the last 40 years has investigated approaches to assessing and training safety skills. This article discusses critical features that have emerged in the research related to best practice for assessing and training safety skills. We emphasize the importance of in situ assessments, effective training approaches, the need for data-based decision making, strategies for enhancing generalization and maintenance, the accessibility of training programs, and approaches to training individuals with disabilities. Following a discussion of each critical feature presented in this article, we provide recommendations for practicing behavior analysts.

A variety of environmental events can lead to the injury or death of children. These safety threats may come from the physical environment or from the behavior of others. Safety threats in the physical environment can include firearm discharge, poisoning, traffic/pedestrian accidents, burns, and drowning (Borse et al., 2008; World Health Organization, 2008). Safety threats from the behavior of others can include abduction and sexual abuse (Miltenberger et al., 2020). Although these safety threats are low incidence such that most children will not experience them, they are highly dangerous. For example, 142 children were killed by firearms in 2020 (Everytown, 2021), over 700 children are killed by poisoning each year (Centers for Disease Control and Prevention, 2021), and over 100 children experience abduction by strangers (stereotypical kidnapping) each year (Office of Juvenile Justice and Delinquency Prevention, 2016). Therefore, parents and caregivers must strive to keep environments safe by eliminating or decreasing the likelihood that children will encounter these safety threats. However, despite parents’ best efforts, children still encounter a variety of safety threats. Therefore, it is important for children to learn the skills to respond safely to these threats. When children exhibit these safety skills, they are more likely to keep themselves safe.

Research has shown that children can engage in safe behaviors in the presence of a safety threat when they are taught specific safety responses (e.g., Miltenberger, 2008; Miltenberger et al., 2020). Safe responses to low-incidence, highly dangerous threats consist of the following three components: (a) identify and avoid the safety threat (i.e., do not touch or engage with it), (b) move away from the safety threat, and (c) report the safety threat to a trusted adult. Discriminating the presence of a safety threat is necessary for the individual to avoid, escape, and report the incident (Miltenberger, 2008). Moving away from the threat is critical because the longer the individual is exposed to the threat, the greater the risk (e.g., Poche et al., 1988). Reporting the safety threat to a trusted adult is critical so the adult can then take action to remove the safety threat. Owing to the importance of these skills for keeping children safe, research over the last 40 years has investigated strategies for teaching and promoting the acquisition and generalization of safety skills with children and individuals with disabilities (Giannakakos et al., 2020; Miltenberger et al., 2020; Tekin-Iftar et al., 2020).

The research on teaching safety skills has investigated a range of interventions for a variety of safety threats. The purpose of this discussion article is to gather from the research the critical features related to successful safety skills instruction and identify practice recommendations based on that empirical research base. The critical features and practice recommendations provided result from a general review of the existing literature and are presented from the perspective of our combined experience conducting related research. To facilitate our efforts in the identification of some of the critical features related to successful safety skills training, we carefully considered the recent literature review by Giannakakos et al. (2020) and reanalyzed the data from relevant studies at the participant level. These features include the importance of in situ assessments compared to other forms of assessment, the effectiveness of different training approaches, the need for data-based decision making, the enhancement of generalization and maintenance, the delivery method of training to increase accessibility, and approaches to training individuals with disabilities. This article will address each of these issues and provide recommendations for practice.

Importance of in Situ Assessment

An assessment is valid when it measures the behavior that it is intended to measure (Kazdin, 2011). As it relates to safety skills, and for the purposes of determining practice recommendations, a valid assessment occurs when a child demonstrates the safety responses in the presence of a simulated safety threat. Valid assessment of safety skills is critical because it is essential in determining the ultimate goal, which is whether the child will engage in the safety skills when faced with a safety threat in the natural environment. Furthermore, because children most often are exposed to safety threats when they are not directly supervised by adults, valid assessment must occur when the child is not in the presence of an adult. Although researchers have described three methods of assessment in the literature—verbal report assessments, role-play assessments, and in situ assessments—in situ assessment is the most valid form of assessment of safety skills (Miltenberger et al., 2020).

Verbal report assessment involves describing a safety threat scenario and asking the child to state how they would respond to that threat (e.g., Gatheridge et al., 2004; Himle et al., 2004b). For example, to assess a child’s knowledge of firearm safety, Gatheridge et al. (2004) described a situation in which the participant found a firearm in the backyard of their home. The researcher asked the participant to describe how they would behave in that situation. Often, children describe safety skills in response to “what if” scenarios, but when presented with a safety threat during an in situ assessment, they fail to engage in the safety skills.

Role-play assessments involve the trainer describing a safety threat scenario and asking the child to demonstrate what they would do (Gatheridge et al., 2004; Himle et al., 2004a; Kelso et al., 2007). For example, the trainer places a disabled firearm on a table and asks the child to pretend they found the gun on a dresser in the parent’s bedroom. The trainer then asks the child to walk into the room, find the gun, and show what they would do. This type of assessment indicates whether the child has the skill in their repertoire. Although role-play assessments can be used to determine whether the child can exhibit the skill in the presence of a safety threat, it is important to note that they do not provide information on whether the child will engage in the safety skills when an adult is not present (Miltenberger, 2008). Research shows that children may exhibit the skills during role-play assessments and then fail to engage in the safety skills during an in situ assessment (Gatheridge et al., 2004; Himle et al., 2004a). The primary issue with role-play assessments is the failure to assess the generalization of the safety skills. For this reason, practitioners should always use in situ assessments to evaluate the effects of safety skills training programs.

An in situ assessment measures the child’s responses to a safety threat in a seemingly real but simulated situation in the absence of an adult without the child’s knowledge (Miltenberger et al., 2020). Researchers have shown that, even when children can describe the skills or demonstrate the skills in a role-play, they do not always perform the safety skills when an adult is absent and the children are unaware of the assessment (Himle et al., 2004a; Miltenberger et al., 1990). Given these findings, it is essential that assessments occur in natural contexts where children are unaware of the assessment and they are not in the presence of a trainer or a trusted adult. For example, Himle et al. (2004a) described how an in situ assessment was conducted to assess safety skills when finding a firearm. During the assessment, the researcher placed a disabled firearm in a room and then asked the child to play in that room. The researchers arranged the assessment so that the child entered the room alone and was unaware that they were being observed. In situ assessments have been shown to be effective for evaluating safety skills with safety threats such as poisons (Dancho et al., 2008; Summers et al., 2011), sexual abuse lures (Lumley et al., 1998; Miltenberger et al., 1999), abduction lures (Beck & Miltenberger, 2009), firearms (Miltenberger et al., 2009), and lighters (Houvouras & Harvey, 2014). If a child does not demonstrate the safety skills during an in situ assessment, it is unlikely that the child will perform the skills when faced with a real safety threat.

When conducting observations during in situ assessments, the use of hidden cameras and mobile devices can facilitate data collection in real contexts. When the use of hidden cameras is not possible, then an alternative is to select a room in the house that allows for multiple discrete vantage points. It is important that the child is unaware that they are being observed so the simulated safety threat seems real to the child. An important consideration when assessing safety skills is to ensure the safety of the child during simulated contexts. For instance, when assessing abduction prevention skills in public, there needs to be ongoing supervision so that the child is never left unobserved during the presentation of a lure. Another consideration is to schedule assessments that occur in community settings during times with fewer customers or visitors (e.g., a local grocery store or playground). Another example is to use a properly disabled firearm (i.e., disabled by someone with proper training and confirmed by a second person) or a nonfunctional replica of a firearm during firearm safety skills training and assessment so the researcher never exposes the child to a firearm that can result in serious harm to the child or others. Similarly, when assessing poisoning prevention skills, researchers must always use innocuous substances (i.e., empty pill capsules) to prevent injury.

Based on the research on in situ assessments, we offer the following recommendations for practice. Behavior analysts should strive to use well-designed, safe in situ assessments to assess the effectiveness of safety skills training because it is the only valid form of assessment to determine whether a child will engage in safety skills in the presence of a seemingly real safety threat in the natural environment. Verbal report assessment and role-play assessment have limited utility for assessing whether the child can state the skills or exhibit the skills when asked. Only in situ assessment is a valid assessment of the generalized use of the skills.

Effectiveness of Training Approaches: What Works and What Does Not?

Research has focused on two approaches to teaching safety skills: informational approaches and active learning approaches. Informational approaches, or passive learning approaches, consist of instructions and modeling without the opportunity for rehearsal and feedback. They have been evaluated to teach several different safety skills to children, such as sexual abuse prevention, abduction prevention, and firearm safety (Beck & Miltenberger, 2009; Gatheridge et al., 2004; Hardy, 2002; Hardy et al., 1996; Miltenberger & Thiesse-Duffy, 1988; Saslawsky & Wurtele, 1986), but none have been found to be effective when the outcomes are evaluated by in situ assessments. Some researchers have evaluated the efficacy of commercially available safety programs such as the Safe Side Stranger Safety program for abduction prevention (Beck & Miltenberger, 2009; Miltenberger et al., 2013) and the Eddie Eagle GunSafe program for firearm safety (Gatheridge et al., 2004; Himle et al., 2004b) and found these programs to be ineffective. In these programs, researchers give children information via instructions and modeling using a variety of learning tools, such as videos, coloring books, posters, and discussions, to teach them what to do if they encounter a safety threat (Beck & Miltenberger, 2009; Carroll et al., 1992; Himle et al., 2004b; Wurtele et al., 1986). However, these programs lack rehearsal and feedback, which have been shown to be critical for teaching safety skills (Gatheridge et al., 2004; Poche et al., 1981, 1988). Children who participate in passive learning programs can describe what they would do in unsafe situations; however, they do not perform the safety skills during in situ assessments (Beck & Miltenberger, 2009; Gatheridge et al., 2004; Himle et al., 2004b; Miltenberger et al., 2013).

Although active learning approaches include both instructions and modeling, the critical features of such programs are as follows: First, the child is given an opportunity to rehearse the skills in the presence of simulated safety threats (the discriminative stimuli [S^Ds] for the safety skills). Second, the child is given praise for correct performance. Third, the child is provided corrective feedback consisting of further instruction for improvement until the child performs the skills consistently (Miltenberger et al., 2020). Behavioral skills training (BST) and in situ training (IST) are two active learning approaches. The literature on teaching safety skills provides substantial evidence for the effectiveness of BST and IST (Giannakakos et al., 2020). BST consists of instructions, modeling, rehearsal, and feedback. During instructions and modeling, the trainer describes the safety threat and the safety skills and demonstrates the safety skills in the presence of simulated safety threats. Thereafter, the child is required to rehearse the skill in the presence of the S^D in a simulated safety context. During rehearsal, the trainer provides praise for steps performed correctly and corrective feedback for steps that are performed incorrectly. Researchers repeat rehearsal and feedback until the child performs the safety skills correctly and independently multiple times.

In an early demonstration of BST, Poche et al. (1988) highlighted the importance of rehearsal and feedback to teach children safety skills. The investigators compared video modeling (VM) plus rehearsal and feedback to VM only, a standard school program, and no-training conditions to teach abduction prevention skills. The performance of the safety skill during VM plus rehearsal and feedback was superior to the other conditions where rehearsal of the skill was not included. Numerous researchers have demonstrated the effectiveness of BST for teaching a variety of safety skills, such as poison prevention, firearm safety, fire safety, abduction prevention, and sexual abuse prevention (Carroll-Rowan & Miltenberger, 1994; Dancho et al., 2008; Gatheridge et al., 2004; Hanratty et al., 2016; Houvouras & Harvey, 2014; Miltenberger et al., 2004). However, this research has shown that BST is not consistently effective for promoting the use of the safety skills with all children during in situ assessments (Gatheridge et al., 2004; Himle et al., 2004a; Jostad et al., 2008; Miltenberger et al., 2004; Novotny et al., 2020). Although BST is a superior approach to passive learning programs, BST is enhanced when it is combined with IST (Giannakakos et al., 2020).

IST consists of BST conducted in the presence of the simulated safety threat in the natural environment. During an in situ assessment, if the child fails to perform the safety skill, the previously unseen trainer intervenes and initiates IST. The assessment is turned into a training session as the trainer requires the participant to rehearse the safety skills multiple times in the natural environment where they encountered the safety threat during the in situ assessment. For example, Gatheridge et al. (2004) compared two training programs to teach 6- and 7-year-old children safety responses when encountering a firearm. They evaluated BST and Level 1 of the Eddie Eagle GunSafe program. The children who failed to perform the safety skills in either group participated in an IST session. In the IST session, the trainer entered the room where the child found the gun and provided instructions, modeling, rehearsal, and feedback. The child had to rehearse the safety skills successfully multiple times in the presence of the “real” gun in the natural environment. IST was effective for participants in the Eddie Eagle group and for participants in the BST group.

A number of other studies have demonstrated the effectiveness of IST after the failure of BST (Himle et al., 2004a; Kelso et al., 2007; Miltenberger et al., 1999, 2004) or in combination with BST (Johnson et al., 2005; Miltenberger et al., 2005). Moreover, some studies have evaluated the efficacy of IST after informational programs failed to teach safety skills (Beck & Miltenberger, 2009; Miltenberger et al., 2013). Researchers have demonstrated IST’s effectiveness with a variety of safety skills, such as abduction prevention (Beck & Miltenberger, 2009; Johnson et al., 2005), sexual abuse prevention (Egemo-Helm et al., 2007), firearm safety (Gatheridge et al., 2004; Miltenberger et al., 2005), and poisoning prevention (Dancho et al., 2008). As cited previously, the majority of studies have evaluated IST following BST, in combination with BST, or following an information-based training program. Few studies have investigated the efficacy of IST as a stand-alone procedure. Miltenberger et al. (2013) compared two groups who received parent-implemented IST. The treatment group watched the Safe Side Stranger Safety DVD, and the control group received no training. After a failed in situ assessment, parents implemented IST with their children. The results showed that IST was effective for the treatment and control groups, suggesting that IST may yield positive outcomes as a stand-alone procedure.

Giannakakos et al. (2020) conducted a systematic review of safety skills training and found that among the 19 experiments that used BST, 47.4% of the studies reported positive results for all participants. Furthermore, among the studies that used a combination of BST and IST, the authors reported that 52.6% obtained positive outcomes for all participants. To assist in determining the critical features and practice recommendations for the current article, we analyzed the studies reviewed by Giannakakos et al. Specifically, we evaluated individual participant data across relevant studies and found that 48.3% of 296 participants who received BST demonstrated the safety skills and 86.3% of 204 participants who received IST demonstrated the safety skills. Of the participants who received IST, it was effective for 89% of 173 neurotypical children and 71% of 31 children or adults with developmental disabilities. Of the participants who received IST, it was effective with 88.4% of 138 individuals who received it after some variation of BST failed, 88.5% of 26 individuals who received it after an informational program failed, and 77.5% of 40 individuals who received it as a stand-alone procedure. These results indicate that when researchers combine BST and IST in a safety skills program, children are more likely to engage in the safety skills during in situ assessments. A noteworthy distinction between data reported by Giannakakos et al. and those included in the current article is that the former researchers based their data analyses on study outcomes, whereas we report data based on individual participants across the relevant studies.

Although BST and IST are effective in almost all cases, research shows these procedures are not effective for some participants (Hanratty et al., 2016; Miltenberger et al., 2004). In cases where IST has not resulted in the generalization of the safety skills to in situ assessments, the addition of tangible reinforcers has proven successful (e.g., Hanratty et al., 2016; King & Miltenberger, 2017; Miltenberger et al., 2004). For example, Miltenberger et al. (2004) showed BST was successful for three of six children and IST for two of three children. However, one child did not engage in the safety skills until the researchers used a tangible reinforcer in training. Unfortunately, researchers do not know in advance for which participants a tangible reinforcer may be required, but it is a reasonable and simple addition should it be necessary.

Other approaches to teaching safety skills that have not received much attention in the safety skills literature are small-scale simulation training and training delivered through virtual reality (VR). Although both approaches include BST, they differ from traditional BST in the way that the training contexts are simulated. Small-scale simulation training involves simulating stimuli from the natural environment into a model and providing training in the model (e.g., Neef et al., 1978, 1989; Page et al., 1976). Maxfield et al. (2019) used a small-scale simulation to teach 3- to 5-year-old children firearm safety skills. Simulation training included all components of BST as the participants manipulated a doll to engage in the safety skills in a tabletop model of a home. After training, the researchers used in situ assessments to assess the performance of the safety skills. Maxfield et al. found that one or two sessions of small-scale simulation training were effective for three of four participants. Orner et al. (2021) replicated the small-scale simulation training with three children with autism but found the procedure worked for only one of the children. Some benefits of using small-scale simulation training are that it is a less expensive approach that can be easily implemented. Although the findings from these two studies are preliminary, they suggest that this approach may be effective in teaching children to engage in safety skills but may be less effective for children with autism.

VR is the use of computer technology to create a real-world context (Çakiroğlu & Gökoğlu, 2019). In VR, the individual is fully immersed in the simulated environment. When using VR, the researcher conducts interactive training in real-life contexts that simulate the safety threat. Another benefit of using VR technology is that safety threats can be simulated virtually for in situ assessments when they would be difficult to simulate in real life for safety threats such as a home fire (Çakiroğlu & Gökoğlu, 2019). Padgett et al. (2006) used a VR game to teach children with disabilities how to respond to a home fire. They used rehearsal and feedback to teach three safety responses similar to those discussed in the studies throughout this article (i.e., identify the fire, avoid the fire, and wait at a designated meeting place). The participants practiced the three responses in the virtual world where praise was delivered for correct responses and corrective feedback was delivered for incorrect responses (e.g., the child walks into a fire). All five participants demonstrated the safety skill in an assessment that consisted of sequencing pictures correctly according to an imaginary scenario. A noteworthy limitation of this study is that the authors did not assess the participants’ responding during an in situ assessment.

In another study, Çakiroğlu and Gökoğlu (2019) taught children fire safety skills (get away from the fire and report the fire to an adult) using BST within the virtual world. Within the VR environment, the authors set up an assessment to observe the participants’ behavior, and when participants did not perform the skills correctly, the researchers implemented IST. They continued this sequence of assessing and providing IST until the participants demonstrated the safety responses. Unlike Padgett et al. (2006), these researchers assessed the generalization of safety responses in an actual environment implemented by a firefighter in a simulated room with a smoke effect and a fire alarm. Of the 10 children who participated in the study, 4 demonstrated the skill during assessments that took place in the virtual world and did not require IST, but only 2 of them performed the safety responses in a real context. Furthermore, six participants did not perform the safety responses in the virtual world, thus requiring IST, which increased their safety score to mastery. However, for two of these six participants, responses did not generalize to the real context. Given the mixed results for effectiveness of BST and IST delivered via VR to teach safety skills to children, further research is warranted.

Based on the research on effective training approaches, we offer the following recommendations for practice. Behavior analysts should use an active learning approach involving BST and IST because it is superior to an informational approach to teaching safety skills and promoting the generalization of the skills to the natural environment. During BST, children must rehearse the skills with feedback until they perform the skills correctly multiple times in the presence of the S^D (the simulated safety threat). In addition, IST should occur in the natural environment to increase the likelihood that the skills will occur in the natural environment posttraining. In some cases, the addition of tangible reinforcers is needed to enhance the effects of BST and IST. Further research is needed to establish the effectiveness of safety skills training delivered via small-scale simulation and VR.

Repeated Assessments and Data-Based Decision Making

Although it is clear that an active learning approach to teaching safety skills is required for success, it is not clear how many BST sessions may be required during training or whether IST or tangible reinforcers may be necessary to promote generalization. Therefore, data-based decision making is necessary when teaching safety skills to determine what intensity of training is necessary for success. For example, when children do not exhibit the safety skills to a specified mastery criterion during in situ assessments following one or two BST sessions, booster training may be required. Himle et al. (2004a); Miltenberger et al. (2004); and Jostad et al. (2008) implemented booster training sessions when BST did not produce generalized responding during in situ assessments. Booster training in these studies consisted of additional BST sessions following failed in situ assessments. Practitioners should implement one or more booster sessions after the initial BST session does not result in the use of safety skills during an in situ assessment. Therefore, to know whether the child needs booster training, the researcher or clinician must conduct in situ assessments and collect data over time to assess the effects of the initial training.

Furthermore, because the safety skills might not generalize to in situ assessments following BST and booster sessions, IST may be needed (e.g., Himle et al., 2004a; Jostad et al., 2008; Miltenberger et al., 2004). Again, because it is not clear in advance whether participants will need IST, repeated in situ assessments are necessary for deciding whether in situ training is necessary. Finally, in some studies, IST was not effective for some children until researchers used a tangible reinforcer in training (e.g., Hanratty et al., 2016; Miltenberger et al., 2004). Researchers could only make the decision to add tangible reinforcers after conducting repeated in situ assessments to evaluate the effects of training. Based on findings from the research, a decision rule for behavior analysts is to move to IST if BST is not effective after three booster training sessions. Finally, if three IST sessions do not produce performance to the mastery criterion, for the purpose of efficiency, behavior analysts should consider using tangible reinforcers.

Based on the research on data-based decision making, we offer the following recommendations for practice. Behavior analysts should use data-based decision making and repeated assessments over time because it is critical for discovering whether additional safety skills training components are necessary for mastery. Because a different intensity of training (i.e., BST, IST, tangible reinforcers) may be required for different children, practitioners need repeated in situ assessments to identify the level of training needed to promote the generalization of the safety skills for a particular child.

Strategies for Promoting Generalization

Assessing and promoting generalization are essential when teaching safety skills to children. Assessing generalization occurs through in situ assessments. Promoting generalization occurs by incorporating generalization strategies during training. Stokes and Baer (1977) described the importance of programming for generalization from the outset to increase the likelihood that the behavior will occur in the relevant environment. Researchers have included various strategies for promoting generalization in their evaluations of safety skills training (Akmanoglu & Tekin-Iftar, 2011; Bergstrom et al., 2014; Godish et al., 2017; Houvouras & Harvey, 2014; Kim, 2016).

In their review, Giannakakos et al. (2020) reported that multiple-exemplar training was used most frequently. Multiple-exemplar training involves including a variety of relevant stimuli within the training context (Stokes & Baer, 1977). If the child learns to respond to a variety of stimuli in different contexts, then it is more likely that they will engage in the safety skills when they encounter similar stimuli in a natural context. For example, when teaching abduction prevention skills, the training program may incorporate different scenarios (e.g., incentive lures or authority lures) and different abductor confederates (Miltenberger et al., 1999). Similarly, a trainer might include different pill containers or place pill containers in different locations when teaching poisoning prevention skills (Morosohk & Miltenberger, 2021; Petit-Frere & Miltenberger, 2020). A second strategy for promoting generalization includes incorporating common stimuli (Stokes & Baer, 1977). The stimuli that are selected for training should come from the natural environment to facilitate the generalization to the natural environment. For example, researchers have incorporated real, disabled firearms when teaching children firearm safety skills (Himle et al., 2004a; Miltenberger et al., 2004). Finally, fading prompts and trainer presence is a strategy that can facilitate generalization when teaching safety skills. Petit-Frere and Miltenberger (2020) and Morosohk and Miltenberger (2021) used generalization strategies with BST to teach poisoning prevention skills to children with autism spectrum disorder. Petit-Frere and Miltenberger included a system of least prompts and a fading procedure in which the trainer moved out of the room when the child rehearsed the skill of reporting the safety threat. Morosohk and Miltenberger added tangible reinforcers to BST and also faded the researcher’s presence during training so the participant had to leave the room to report the pills to an adult. In both studies, all participants exhibited the safety skills during in situ assessments following one or two BST sessions.

Another generalization strategy that behavior analysts may consider is mediating generalization by having children describe the skills that they are performing. By describing the safety skills as they exhibit them during training sessions, the skills might come under the stimulus control of the verbal statements or rules. Unfortunately, little research has evaluated this generalization promotion strategy. Olsen-Woods et al. (1998) evaluated correspondence training (CT) and found it did not enhance the effects of BST. Other researchers had participants describe the skills as they exhibited them, but did not evaluate the additive effects of this added intervention component (e.g., Maxfield et al., 2019; Page et al., 1976). Thus, more research is needed to evaluate the value of this generalization strategy.

The aforementioned strategies included generalization tactics within BST and demonstrated successful generalization of the skills to in situ assessments. However, most of the studies evaluating BST to teach safety skills have added IST to enhance generalization (Godish et al., 2017; Himle et al., 2004a; N. Lee et al., 2019; Miltenberger et al., 2004; Morgan & Miltenberger, 2017; Novotny et al., 2020). Numerous safety skills studies indicate that IST may be necessary following BST to produce the generalization of the safety skills to in situ assessments (e.g., Bergstrom et al., 2014; Himle et al., 2004a; Miltenberger, 2008). When researchers added IST after a failed in situ assessment, children were more likely to engage in the safety skills (Himle et, 2004a; Miltenberger et al., 2004). This research demonstrates that training in a simulated but realistic safety threat situation is highly effective for promoting generalization. However, IST typically is implemented following the failure of BST and thus takes more time and effort. Therefore, it is desirable to incorporate strategies into BST to promote generalization so that the addition of IST may not be necessary.

Based on the research on promoting generalization, we provide the following recommendations for practice. Implementing strategies to promote generalization is essential when teaching safety skills because generalization of the skills to in situ assessments is the only valid indication of success. Behavior analysts should incorporate generalization strategies into BST and assess the effects of training with in situ assessments. At a minimum, practitioners should use multiple-exemplar training and incorporate common stimuli to enhance generalization. Practitioners should use IST as a default approach to promoting generalization when BST that incorporates generalization-enhancing procedures is not effective.

Importance of Assessing and Promoting Maintenance

Evaluating the maintenance of training outcomes for safety skills is important for assessing the durability of the acquired skills across time. Much of the safety skills literature includes a measure of maintenance following intervention (Giannakakos et al., 2020). Based on our reanalysis of relevant studies from Giannakakos et al. (2020) at the participant level, 72% incorporated follow-up assessments. Furthermore, 42% of these studies achieved success during all maintenance probes for all participants, and 55% of studies achieved success for at least one participant during one or more maintenance probes. Finally, only one study did not obtain positive results during maintenance probes for any participant (Lumley et al., 1998). Among the studies wherein only some participants performed correctly during follow-up probes, 21% conducted IST following a failed assessment (e.g., King & Miltenberger, 2017; Miltenberger et al., 1999). In eight studies that incorporated IST when a participant failed to perform the safety skills during follow-up consistently, the participants engaged in the safety skills in subsequent assessments. For studies that demonstrated partial success during maintenance probes, participants demonstrated improved skills when compared to performance during baseline. Furthermore, among the 55% of studies that achieved some success, the participants maintained the safety skill during at least one follow-up interval (e.g., a participant in Jostad et al., 2008, maintained the safety skills during the 6-month follow-up session but not during the 11-month follow-up session). Given that the studies that included IST following a failed maintenance probe all yielded effective results, all safety skills training programs should use IST after a failed maintenance assessment to achieve optimal outcomes.

We analyzed the data from several studies to get a picture of the length of maintenance evaluations in research. The length of follow-up probes varied across studies; 10.5% of studies assessed maintenance at between 1 and 3 weeks (e.g., Miltenberger et al., 2013; Petit-Frere & Miltenberger, 2020), 73.7% of the studies assessed maintenance at between 1 and 3 months (e.g., Beck & Miltenberger, 2009; Dancho et al., 2008; Egemo-Helm et al., 2007), and 15.8% of studies assessed maintenance at between 4 and 12 months (e.g., Jostad et al., 2008; Miltenberger et al., 2004; Tarasenko et al., 2010). No published studies have assessed postintervention outcomes beyond 12 months.

Incorporating rules into training may be useful for increasing the likelihood that the safety skills will maintain. Olsen-Woods et al. (1998) evaluated CT to teach children abduction prevention skills. Children in the training-with-CT group received reinforcement following a verbal response related to the safety skill and performance of that skill, whereas the children in the training-without-CT group were required to perform the skill without providing a verbal response. The researchers found that CT did not increase correspondence between the statements made by the participants and their performance. In fact, both groups performed similarly during in situ assessments. Therefore, more research is required to evaluate the role that rules have when assessing the maintenance of skills.

Based on the research on maintenance, we offer the following recommendations for practice. Behavior analysts should assess the maintenance of safety skills over time to ensure that children engage in safe behaviors in the event that they encounter a safety threat months after training. Although many studies include follow-up assessments, very few assess long-term outcomes (e.g., longer than 3 months). Behavior analysts should conduct in situ assessments to assess maintenance beyond 3 months and should be prepared to conduct IST if needed at each follow-up assessment.

Accessibility of Training

An abundance of research has shown that BST and IST are effective for teaching safety skills to children. However, these procedures are time consuming and resource intensive as a behavior analyst with specific training in teaching safety skills is needed to carry out the procedures (e.g., Miltenberger, 2008; Novotny et al., 2020). These potential barriers may limit the adoptability of effective procedures for teaching safety skills (Jostad & Miltenberger, 2004). As a result, researchers have evaluated different delivery methods that may increase the widespread adoption of such practices. Researchers have addressed increasing accessibility by evaluating training conducted by parents, teachers, and peers and by evaluating simulation training.

Gross et al. (2007) developed a training manual and video to teach parents to conduct BST and IST with their children. The results showed that three out of the four parents implemented the procedures with fidelity and their children engaged in the safety skills during in situ assessments. Furthermore, the authors reported that the total parent-training time, including preparing for sessions, was less than 1 hr. In another parent-training study, Novotny et al. (2020) evaluated web-based training for parents to conduct BST to teach firearm safety skills to their children. The website they developed included a series of videos that incorporated all of the BST components. For children who did not perform all three target behaviors after parents conducted training, a researcher conducted IST. Three of the six children demonstrated the safety skills following parent-conducted BST, and three children required experimenter-implemented IST. This study reported high treatment integrity during training with a mean of 95.5%. These findings lend further support for parent involvement in training children to engage in safe behaviors when in the presence of a firearm.

Other researchers have also evaluated parent-implemented IST to teach their children safety skills (Beck & Miltenberger, 2009; King & Miltenberger, 2017; Miltenberger et al., 2013; Morgan & Miltenberger, 2017). For example, Beck and Miltenberger (2009) and Miltenberger et al. (2013) used BST to teach parents to conduct IST to teach abduction prevention skills after an informational program was ineffective. In these studies, most parents implemented the procedures with fidelity. However, in some cases, researchers had to implement IST when parents did not implement procedures with fidelity (King & Miltenberger, 2017; Morgan & Miltenberger, 2017).

In addition to successfully training parents to implement BST and IST, researchers have evaluated procedures for training teachers to implement BST with their students (Gast et al., 1993; Holcombe et al., 1995). Carroll-Rowan and Miltenberger (1994) evaluated two training procedures to teach abduction prevention skills to children in a classroom setting. The researchers found that, during in situ assessments, children in the training groups performed the safety skills whereas children in the control group performed lower. That study demonstrates that children can be taught abduction prevention skills in a classroom setting when training is delivered by their teachers.

Another approach for increasing efficiency with safety skills training has been to incorporate peer trainers (Jostad & Miltenberger, 2004). Researchers trained a few children to conduct BST and IST and the children then conducted the trainings with other children. Jostad et al. (2008) used 6- and 7-year-old peer trainers to teach 4- and 5-year-old children firearm safety. The peer trainers taught students individually using instructions, modeling, rehearsal, and feedback. Following training, if the student failed to engage in the safety skills, a peer trainer implemented IST. The findings show that peer tutors can be effective in teaching safety skills to younger children. Furthermore, Jostad et al. reported that peer-implemented BST required less than 2 hr to complete with little adult assistance. This is an important finding because if training can be done by peers in less time and is as effective as when implemented by an adult trainer, then peer-implemented BST can have far-reaching outcomes for children. In a similar study, Tarasenko et al. (2010) evaluated peer training for teaching abduction prevention skills to children. Their results were similar to those of Jostad et al., demonstrating that peer trainers effectively used BST and IST to teach the safety skills in the presence of abduction lures. The authors reported that training the peer trainers lasted 1 hr and training the students ranged from 18 to 25 min. These two studies support the efficacy of using peers to conduct evidence-based practices for teaching safety skills to children.

Although research shows that parents and teachers can be taught to teach safety skills to children, no research has been conducted yet teaching behavior analysts to teach safety skills. Given the success with parents and teachers, researchers should evaluate strategies for teaching behavior analysts to conduct safety skills training with a focus on web-based approaches that are efficient and accessible.

Another intervention that might contribute to the increased accessibility of training is VM. Although researchers have combined VM with behavioral rehearsal and feedback to teach safety skills (Carroll-Rowan & Miltenberger, 1994; Poche et al., 1988), they have also evaluated VM as a stand-alone procedure. Godish et al. (2017) evaluated the effectiveness of VM to teach abduction prevention skills to children with autism spectrum disorder. The videos used during training depicted a child model who was approached by a stranger and said no, ran away, and reported the safety threat to an adult. The video also had an interactive component in which the video paused after the child in the video was presented with an abduction lure and the narrator asked the viewer what the child should do. The viewer had a chance to respond vocally, and then the narrator described the correct answer (“If you said the boy should say no, get away, and tell an adult, you are right!”). The findings from this study showed that VM alone was effective for all four participants, although one needed IST at follow-up. Furthermore, Akmanoglu and Tekin-Iftar (2011) evaluated VM as part of a treatment package that included graduated guidance to teach children with autism spectrum disorder how to respond to lures from a stranger. The results of this study showed that VM was effective for teaching abduction prevention skills to all three participants, although the researchers included a prompt and prompt-fading procedure that may have contributed to the success of VM.

Despite the success of these two studies, other studies showed VM was not effective (King & Miltenberger, 2017; Morgan & Miltenberger, 2017). King and Miltenberger (2017) found VM to be ineffective for teaching poison safety skills to all three participants. Similarly, Morgan and Miltenberger (2017) reported success with firearm safety skills for only one of three participants. Both studies incorporated IST following failed in situ assessments resulting in performance to the mastery criterion for all but one participant. Given the mixed results obtained by studies evaluating VM, additional research is needed to determine the effectiveness of this training approach.

A final training approach that might increase accessibility is computer simulation training. Vanselow and Hanley (2014) evaluated the effects of a computerized version of BST (CBST) to teach children safe responses to different safety threats. The researchers evaluated the computerized safety skills training with in situ assessments and conducted IST if necessary. Through a series of studies, they assessed the generalization of the skills to threats that were not explicitly trained, such as poisons and strangers. The authors found that all participants acquired the safety responses in the presence of all three safety threats, but generalization occurred with only specific threats. Finally, not all participants demonstrated the safety skills after CBST alone; some required the addition of IST to perform the skills. Given the ease of implementation and accessibility of computerized safety skills training programs, further research is warranted to evaluate its success. Another area of research within safety skills training that has not received attention is telehealth. Due to a variety of constraints that may prevent in-person training, telehealth seems to be a viable option. However, research is needed to evaluate the efficacy and efficiency of this training approach for teaching safety skills.

Based on the research on accessibility of training, we offer the following recommendations for practice. Behavior analysts should strive to increase the accessibility of successful training procedures. Studies show that teachers, parents, and peers can teach safety skills successfully when trained by researchers. This training can be through direct contact using BST or through training materials the individuals can access independently. More research is needed to establish the robustness of these procedures. Furthermore, researchers should evaluate web-based procedures for teaching behavior analysts to conduct safety skills training. If behavior analysts train others to conduct safety skills training, it is important to assess implementation fidelity to assure competent performance before they conduct training. One other promising approach is simulation training—both small-scale simulation and computer simulations. However, more research is needed to establish the success of these procedures.

Training Individuals With Disabilities

Although all children are at risk of harm from safety threats, children with developmental disabilities are at greater risk for harm (L. Lee et al., 2008). This increased risk may be due to behavioral deficits such as difficulty with communication and social interactions and limited intellectual abilities that are characteristic of developmental disorders (Dixon et al., 2010). Furthermore, researchers have reported that perceived deficits of safety skills may result in parental overprotection, potentially limiting the development of independent skills (Wiseman et al., 2017). Although parents and professionals have reported safety skills to be a primary concern for children with developmental disabilities (Collins et al., 1991), they often focus on skills that require immediate attention, such as daily living skills or communication skills, and place less emphasis on safety skills during instructional contexts (Dixon et al., 2010; Tekin-Iftar et al., 2020). Furthermore, when it comes to safety skills, researchers have reported that parents and educators simply provide warnings of safety threats or arrange the environment so that safety threats are not present (Tekin-Iftar et al., 2020). Although it is important that adults ensure safety threats are eliminated from the environment, children with and without developmental disabilities need to be taught safety skills in case they encounter safety threats. However, because children with developmental disabilities have more limited repertoires, more intensive training may be required.

Numerous studies show that BST and IST are effective with neurotypical children, but fewer studies have evaluated interventions to teach safety skills to children and adults with developmental disabilities. Researchers have evaluated BST and IST for skills such as sexual abuse prevention (Egemo-Helm et al., 2007; Haseltine & Miltenberger, 1990; Lumley et al., 1998; Miltenberger et al., 1999), abduction prevention (Akmanoglu & Tekin-Iftar, 2011; Bergstrom et al., 2014; Fisher et al., 2013; Godish et al., 2017; Gunby & Rapp, 2014; Gunby et al., 2010; Ledbetter-Cho et al., 2016; Sanchez & Miltenberger, 2015), poisoning prevention (King & Miltenberger, 2017; Petit-Frere & Miltenberger, 2020; Summers et al., 2011), and avoidance of fire-starting agents and firearms (Morgan & Miltenberger, 2017; Rossi et al., 2017). Although research shows that BST and IST (with the occasional need for tangible reinforcers) are effective for neurotypical children (Hanratty et al., 2016; Himle et al., 2004a; Johnson et al., 2006; Jostad et al., 2008; Miltenberger et al., 2004, 2005), a larger range of procedures has been evaluated for teaching safety skills to individuals with developmental disabilities. We found approaches to training safety skills to individuals with developmental disabilities fall into the following categories: (a) BST alone, (b) enhanced BST, (c) BST with complementary interventions, (d) and BST enhancements and complementary interventions. Among the studies we considered, 82% incorporated BST in their interventions with participants with developmental disabilities. However, only one study evaluated BST alone with individuals with developmental disabilities (Lumley et al., 1998). Lumley et al. (1998) investigated a sexual abuse prevention program and compared different assessments methods: verbal report, role-play, and in situ assessments. Although the participants demonstrated the safety skills during role-play assessments after BST, they did not demonstrate the skills during the in situ assessments. Consistent with the literature, the researchers reported a lack of correspondence between the different assessment methods with individuals with developmental disabilities.

Among the studies we considered that used BST to train safety skills, 39% enhanced BST to increase its effectiveness. Enhancements to BST can be defined as procedures that are embedded within training, rather than adding interventions after a failure of BST. Prompting and prompt-fading procedures, tangible reinforcers, and VM are all examples of procedures embedded in BST that have produced positive outcomes for training children with developmental disabilities (Bergstrom et al., 2014; Ledbetter-Cho et al., 2016; Petit-Frere & Miltenberger, 2020; Summers et al., 2011). Petit-Frere and Miltenberger (2020) evaluated BST with the system of least prompting to teach poisoning prevention skills to children with autism. The procedure was effective for all three children. Morosohk and Miltenberger (2021) enhanced BST by adding tangible reinforcers for correct responding to teach poisoning prevention skills to children with autism spectrum disorder. During in situ assessments, all participants demonstrated the safety skills following one or two training sessions.

Researchers have shown that complementary interventions, consisting of the addition of IST and incentives to BST following a failed in situ assessment, can be effective when training individuals with developmental disabilities (Gunby & Rapp, 2014; Haseltine & Miltenberger, 1990; Sanchez & Miltenberger, 2015). Based on our general review of related studies, we found 28% of studies added complementary interventions to BST resulting in treatment gains for all participants. For example, Egemo-Helm et al. (2007) added IST when BST was not effective for promoting the use of sexual abuse prevention skills during in situ assessments for participants with developmental disabilities. In another study, Sanchez and Miltenberger (2015) found similar results implementing BST and IST to teach abduction prevention skills to adolescents with developmental disabilities.

Other researchers developed training programs that include both enhancements to BST and complementary strategies when teaching individuals with developmental disabilities. Among the studies we reviewed, 17% modified an aspect of BST during initial training and added intervention components. For example, in a study evaluating a sexual abuse prevention program, Miltenberger et al. (1999) enhanced BST by providing tangible reinforcers following a correct response when participants were presented with a sexual abuse lure. When modified BST did not result in criterion performance during in situ assessments, the researchers conducted IST. Miltenberger et al. (1999) showed that IST was effective in producing generalized responding for four of the five participants. Other researchers have included incentives when they did not obtain positive outcomes following IST (Godish et al., 2017; King & Miltenberger, 2017; Ledbetter-Cho et al., 2016). In another example, Ledbetter-Cho et al. (2016) used modified BST that included a video model, rather than a live model, to teach four children with autism to respond to multiple types of abduction lures. For one participant who did not reach the performance criterion during posttraining assessments, the researchers included a contingency-specifying instruction with reinforcement. This participant reached the mastery criterion for two out of the four types of lures during posttraining assessments.

Based on research with individuals with disabilities, we offer the following recommendations for practice. Because the generalization of safety skills to in situ assessments may be more problematic for individuals who have limited repertoires (Gunning et al., 2019), behavior analysts should incorporate strategies that will facilitate generalized responding when teaching safety skills to children or adults with developmental disabilities. Research has shown that training is more likely to be successful when enhancements are made to BST (e.g., tangible reinforcers, prompting strategies, stimulus control strategies) or when complementary interventions (e.g., IST) are included. Behavior analysts should take into account the differences in skill level and the need for enhancements when training persons with developmental disabilities.

Summary

This article presented a number of critical features in the research on teaching safety skills that behavior analysts need to consider to teach safety skills successfully. We also presented recommendations for practice relevant to each issue. A summary of our recommendations is as follows:

1. Use in situ assessments to evaluate the effects of safety skills training procedures.

2. Use active learning approaches to training, including BST and IST, which involve repeated rehearsals of the skills in the presence of simulated S^Ds with feedback.

3. Use data-based decision making with repeated in situ assessments to determine the intensity of training needed for success with each individual.

4. Implement strategies in training to promote the generalization of the skills to in situ assessments.

5. Assess the safety skills at regular intervals following training and be prepared to implement IST if needed.

6. Strive to increase the accessibility of training by training others, such as teachers, parents, staff, and peers, to carry out safety skills training procedures.

7. Modify training procedures to ensure success for individuals with autism and developmental disabilities.

Authors’ contributions

Both authors contributed to conceptualization and writing.

Funding

There was no funding for this research.

Declarations

Conflicts of interest

None.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Both authors consent to publishing this article.