In 2004–2005, 71,000 children were taken to the emergency room in the USA as a result of accidental poisoning, and 80% of these cases were due to an unsupervised child finding and consuming medication (Centers for Disease Control and Prevention [CDC] 2012). The consumption of medications is two times more common than poisonings resulting from household cleaning products, leading to more emergency hospital visits (CDC 2012). In addition, children under 6 years of age are the most vulnerable population for accidental poisoning (CDC 2012).

Preventing children’s access to such materials is the first step in reducing the number of accidental poisonings among children. Two studies evaluated parent training to decrease children’s access to dangerous items in the home (Barone et al. 1986; Lutzker et al. 1998). Barone et al. (1986) gave parents materials (such as latches and locks) to attach to cabinets and had them watch videos on how to install those devices and on how to keep dangerous objects (such as poisonous solids and liquids) out of their child’s reach. Lutzker et al. (1998) used instructions, modeling, practice with feedback, and homework assignments to teach parents to reduce the number of accessible hazards in the home. The results showed that these procedures reduced but did not eliminate hazards in the homes.

Behavioral skills training (BST) and in situ training (IST) are active training approaches that have been demonstrated to be effective for teaching safety skills to children (e.g., Gatheridge et al. 2004; Himle et al. 2004; Miltenberger 2008). BST involves delivering instructions, modeling the correct target behaviors, allowing the participant to rehearse the target behaviors in a variety of different role play scenarios, and providing the participant with praise or corrective feedback based on his or her performance. IST involves the trainer entering into an in situ assessment and conducting on-the-spot training if the participant does not engage in the target behaviors in the presence of the safety threat (Miltenberger 2008).

Video modeling combined with behavioral rehearsal and feedback is another effective training approach for teaching safety skills to children (Carroll-Rowan and Miltenberger 1994; Poche et al. 1988). Recent studies have also evaluated video modeling exclusively and video modeling combined with other active learning procedures for teaching safety skills to children diagnosed with autism spectrum disorder (ASD) (Godish et al. 2017; Gunby et al. 2010; Morgan and Miltenberger 2017). Godish et al. (2017) used video modeling to teach abduction prevention skills to four children (ages 7–8) diagnosed with ASD. All the participants engaged in the target behaviors following video modeling, although BST and IST were needed for one participant during the follow-up assessments. Morgan and Miltenberger (2017) tested the effectiveness of video modeling to teach firearm avoidance skills to three 6 year olds diagnosed with ASD. Video modeling was only effective for one participant, however, and the other two acquired the three target behaviors (do not touch, leave the room, and tell an adult) following IST implemented by a parent.

The results of Godish et al. (2017) and Morgan and Miltenberger (2017) show that for children diagnosed with ASD, video modeling was effective for teaching safety skills associated with a social safety threat (an abduction lure), but was not effective for teaching safety skills related to a physical safety threat (guns). More research should be conducted with video modeling to evaluate the specific safety threats, behaviors, and populations for which it is effective in teaching safety skills. Considering the effectiveness of video modeling for teaching social, play, language and communication, self-care, and daily life skills to children diagnosed with ASD (e.g., Acar and Diken 2012; Bellini and Akullian 2007; Charlop-Christy et al. 2000) and the mixed results from studies evaluating video modeling for teaching safety skills, there is a need for more research evaluating video modeling on a variety of safety skills for children with ASD. Therefore, the purpose of this study was to evaluate the effectiveness of video modeling to teach poison prevention skills to three children diagnosed with ASD.

Method

Participants and Settings

This study included three 6-year olds diagnosed with ASD recruited through a local ASD clinic. All three participants received applied behavior analysis (ABA) services prior to and during the course of their participation in this study. The researcher completed a brief assessment to determine whether the child had the prerequisite skills to participate in the study. The researcher asked the participant to complete multi-step instructions independently, deliver a five-word to six-word message to another person after walking 3 m, and label an item or activity that he or she was engaged in 30 s after the item or activity had been terminated. Baseline assessments, video modeling, in situ training, and follow-up assessments occurred in the behavior clinic (exclusively for Jill) or both the behavior clinic and the participants’ home (for Carl and Nicole).

Jill was an only child and lived with her parents. She was diagnosed with Autism at age 2. She also had diagnoses of sensory integration disorder, anxiety, eosinophilic esophagus, hypotonia, and allergic rhinitis. She was in first grade and attended a public school in a special education class. She received 90 min a week of speech therapy in the school setting as well as speech therapy, physical therapy, and occupational therapy in a private setting. She received 15 h a week of ABA therapy after school. Jill had a strong mand, tact, and echoic repertoire. She typically spoke using three-word utterances, but could imitate five-word to six-word phrases when asked. Jill was on the following medications throughout all phases of the study: focalin, zoloft, zyrtec, and elemental formula (Neocate EO28 Splash). Jill’s mom reported that Jill was used to seeing and taking her own medication but that these medications were never readily accessible to Jill and it was unknown how Jill would respond to finding medication that did not belong to her.

Carl and Nicole were fraternal twins who attended kindergarten in a public school. Carl was in a typical class, while Nicole was in a special education class. Both children were diagnosed with Autism; Carl was diagnosed at age 3 and Nicole at age 2. Carl had an IQ score of 96, and Nicole had a score of 65 with a score of 85 for the nonverbal portion of the test and a score of 55 for the verbal portion of the test. They lived with their parents and received two 3-h sessions of ABA therapy a week after school in the clinic setting. They also received 90 min a week of speech therapy in school. Both children had a well-established mand and tact repertoire. Carl had a more advanced verbal skills and typically spoke using five to six words in complete sentences. Nicole typically spoke using one-word to two-word mands, but could demonstrate tacts, five-word to six-word echoics, and intraverbal skills when probed. Carl and Nicole were not taking any medications during the course of this study. Mom reported that both children were only exposed to taking medications if they were sick or were visiting family members who regularly took medication and left them in a visible location. It was unknown how the children would respond to finding medication that was not their own in both the clinic and home setting.

Materials

Videos

This study included three videos (5 to 6 min each) the researchers made based on the research of Morgan and Miltenberger (2017). Each of the three videos included six scenarios showing a child model coming into contact with a poison hazard (pills), engaging in the target behaviors, and receiving praise from an adult for engaging in the target behaviors. Three different forms in which pills may typically be found (a prescription bottle, pill box, and Ziploc bag) were portrayed in two scenarios for each of the three videos. The pills were placed in various locations around the house (kitchen, bathroom, living room, coffee table, parents’ bedroom, child’s bedroom, pool deck) and school setting (bathroom, picnic table, hallway, materials room, desk, table, front office) to promote generalization. Three different models were used in the video to promote generalization. Models for the videos included a typically developing 10-year-old boy and 7-year-old girl, and a 7-year-old boy diagnosed with ASD.

The first video included a verbal prompt to attend to the model’s behavior for all six scenarios. For example, at the beginning of each scenario, the narrator said, “Watch closely as Anthony walks into the bathroom and finds a bottle of pills. He does not touch it, leaves the room right away, and tells an adult about it.” The second video included the verbal prompt for the first three scenarios, but not for the following three. The third video did not include the verbal prompt for any of the scenarios.

Each scenario showed the child walking into a room in which pills were present and included an interactive component in which the scene paused and the narrator asked the participant what the child should do, such as “Anthony just walked into the bathroom and saw a bottle of pills on the sink. What should he do?” The scene continued to pause for 10 s to allow for a response from the participant. After 10 s, the narrator continued, “If you said Anthony should not touch the bottle, leave the room, and tell an adult, you are right! Let’s watch.” If the participant did not respond to the narrator’s question within 5 s or the participant responded incorrectly, the researcher prompted the child to emit the correct response. After the narrator provided praise and feedback, the model engaged in the target behaviors and received praise from an adult for reporting the presence of pills.

Poison Materials

White and colored empty pill capsules were used in this study to mimic actual medication. The use of empty pill capsules instead of small pill-like candies eliminated the possibility that ingesting candies resembling pills during assessment would be inadvertently reinforced. Pill containers included prescription bottles, pillboxes, and clear Ziploc bags.

Target Behaviors and Data Collection

Three safety skills were assessed when the participants came into contact with the pill container: do not touch it, leave the room, and tell an adult. The target behaviors were scored on a four-point scale: 0 = touched the pill container; 1 = did not touch the pill container, but did not leave the area within 10 s of seeing the pills; 2 = did not touch the pill container, left the area within 10 s of seeing the pills, but did not tell an adult about the presence of the pills; and 3 = did not touch the pill container, left the room within 10 s upon seeing the pills, and told an adult about the presence of the pills; a score of 3 was given. The following behaviors were also recorded if they occurred: attempted to open the container, opened the container, touched the pills, removed the pills from the container, and put the pills in the mouth.

Frequency data were collected on the following occurrences during video modeling: the number of prompts delivered to the participant to attend to the video, the number of correct answers emitted by the participant, the number of prompts to evoke the correct answer, and the number of incorrect answers to the narrator’s questions during the interactive component of the videos.

For IST, data were recorded on the number of incorrect and correct performances during the session and whether the participant engaged in three consecutive correct performances before the session was terminated.

Assessment

In situ assessments were conducted during all phases of the study to assess the participants’ safety skills. Assessments were approximately 1 min and were conducted in the behavior clinic, the participants’ homes, or both. In the home, the parent asked the child to do something in the room or location in which the pill container was placed (e.g., to play with toys in a toy chest where a pill box was placed beside or on the toy chest). If the assessment took place in the clinic, the therapist working with the participant asked the child to go to the room or location where the pill container was placed. Without the child’s knowledge, a pill container was placed in an obvious location prior to the assessment. Each assessment was different from the previous assessment in location and reason for sending the child to the area. The participants were kept unaware of the researcher’s presence to reduce stimulus control effects. A video monitor and camcorder were set up for inconspicuous observation of the participants’ responses during in situ assessments.

Interobserver Agreement

Interobserver agreement (IOA) data were collected for 33% of the in situ assessments during all phases of the study. IOA data were also collected for the behaviors and events recorded during the video modeling and IST sessions. The PI and research assistants were responsible for collecting data, with one being the primary and one being the reliability observer. IOA data collection occurred live during sessions or by watching video recordings of the sessions. Agreement was calculated by dividing the number of agreements that each target behavior did or did not occur by the number of agreements plus disagreements then multiplying by 100.

The percentage agreement on the occurrence of the safety skills was 98% (range 87 to 100%) for Jill and 100% for Carl and Nicole. For video modeling, percentage of interobserver agreement on the target behaviors while watching the video was 100% for Jill, 82% (75% to 88%) for Carl, and 100% for Nicole. Percentage agreement on the target behaviors occurring during IST was 93% (range 70% to 100%) for Jill and 100% for Carl and Nicole.

Side Effects and Social Validity Questionnaire

A social validity questionnaire adapted from Johnson et al. (2005) was distributed to parents after their child had mastered the safety skills. The questionnaire assessed parental observations of behavior changes in their children, parental attitudes regarding the training process, parental preference of intervention methods, and parental perceptions on the importance of teaching poison prevention skills to children (see Appendix).

Research Design

A multiple baseline design across participants was used to evaluate video modeling for teaching poison prevention skills. If video modeling was not effective, IST was conducted to teach the poison prevention skills. Mastery criterion was met when the participant scored a 3 for three consecutive assessments.

Treatment Fidelity

Treatment fidelity data were recorded for 33% of sessions for both video modeling and IST to ensure that the procedures were implemented with integrity. For video modeling, treatment fidelity data recorded if the researcher accurately provided a prompt for the child to respond to the narrator’s question within 5 s if the child failed to do so, provided prompts until the child said the correct response (“don’t touch, get away from the pills, tell an adult”), and provided praise when the child emitted the correct response if prompts were delivered. Treatment fidelity for video modeling was 100% for Jill, Carl, and Nicole.

For IST, treatment fidelity data were recorded to ensure that the researcher accurately implemented IST by (a) instructing the child to not touch the pill container, leave the area within 10 s, and tell an adult about the presence of the pill container; (b) having the child practice the skills until correct three times; and (c) providing praise for correct performances and corrective feedback for an incorrect performances. Treatment fidelity for IST was 100% for all participants.

Procedures

The researcher met with the parents to discuss the purpose of the research study, the procedures, the logistics of the training process and time commitment, as well as the child’s verbal repertoire. After meeting with the parents, collecting the consent forms, and completing the informal probe assessment, baseline was implemented.

Baseline

In situ assessments were completed with each participant until a stable level of data was achieved. No consequences or feedback were delivered in this condition. However, if the child reported the presence of the pill container to an adult, praise was delivered.

Video Modeling

The researcher met with each participant three times in 1 week to have the participant watch one of the three videos for each session until the participant watched all three videos. The researcher was responsible for ensuring that the participant responded correctly within 5 s to the narrator’s questions. After the narrator posed a question, the scene paused and the researcher prompted the participant to respond if he or she did not answer the question within 5 s or answered incorrectly. If the participant responded correctly, the video continued to play and the narrator provided praise. If the participant did not respond correctly, the researcher prompted the correct response and repeated the question until the participant emitted the correct response. Once the participant said the correct response, the researcher provided praise and continued the video.

Within 1 h of watching the video (but no earlier than 45 min), the parent or therapist asked the child to go to a room or location where the pill container was placed in an obvious location. The participant’s responses were observed via a hidden video monitor. If the participant scored a 3, the parent or therapist provided praise immediately upon the child informing them about the presence of the pill container. These procedures (presenting the video and then conducting an in situ assessment within 1 h) were repeated two more times on separate days. If the participant did not score a 3 for three consecutive assessments following the implementation of all three videos, IST was implemented.

In Situ Training

The researcher or a trained research assistant was responsible for implementing IST. If the child did not engage in all three target behaviors during the assessment, the researcher entered the room and turned the assessment into an IST session. In an IST session, the researcher instructed the child to not touch the pills, leave the area within 10 s, and report the presence of the pills to his or her parent (if sessions took place in the home) or to his or her therapist (if sessions took place in the clinic). The researcher had the participant practice the target behaviors in a variety of different scenarios (e.g., finding a pill box on the bathroom sink, finding pills in a Ziploc bag in a toy box, finding pills in a prescription bottle on the participants’ bed). After each scenario, the parent or therapist provided praise if the participant engaged in all of the target behaviors and the researcher provided corrective feedback if the participant engaged in some target behaviors incorrectly or not at all. In situ training was conducted until the participant received a score of 3 for three consecutive trials, and then, another in situ assessment was scheduled within 1–3 days. Target behaviors were considered mastered when a participant scored a 3 for three consecutive in situ assessments.

Incentive Condition

If participants did not meet mastery criteria following IST, an incentive condition similar to the one conducted by Miltenberger et al. (2004) was implemented in which the participant had access to a preferred tangible (5 min of computer) contingent on the execution of all three target behaviors in an assessment. The participant was exposed to the incentive condition during a priming session, so she had experienced the contingency prior to an in situ assessment.

In a priming session, the researcher explained the contingencies for engaging and not engaging in the target behaviors to the participant. For example, the researcher told the participant that if she engaged in the target behaviors, then she would have access to computer. In addition, the researcher told the participant that if she did not engage in the target behaviors, then she would not have access to computer. The researcher then sent the participant to a location where the pill container was placed. If the participant engaged in the target behaviors, she earned access to computer for 5 min. If the participant did not engage in the target behaviors, she was told that she did not earn computer and to try again. After the participant engaged in the target behaviors and earned access to computer two to three times, the priming session was terminated and an in situ assessment was conducted within 1 h to a day later.

The incentive condition was used for Jill, and the reinforcer was 5 min of computer access. Access to the computer was chosen because it was reported by Jill’s ABA therapists to be a potent reinforcer during her therapy sessions and was more preferred than attention or access to other tangibles.

Follow-Up Assessment

In situ assessments were conducted in 1, 3, and 5-week intervals after participants mastered the skills in training to assess for maintenance of the skills. If the participant did not score a 3, IST was implemented and another assessment occurred within 1–3 days.

Results

Results for all participants are depicted in Fig. 1. Video modeling was not effective in teaching the safety skills to any of the participants, but IST was effective for two of the participants. Carl and Nicole mastered the skills with IST and maintained the skills at their 1, 3, and 5-week follow-up assessments. Jill required an additional incentive condition before mastering the skills but did not maintain the skills in her 1-week follow-up assessment. However, after receiving two priming sessions, she scored a 3 and maintained a 3 at the 3 and 5-week follow-up.

Fig. 1
figure 1

The safety skill score across phases for Jill, Carl, and Nicole. Clinic data are depicted with circles, and home data are depicted with triangles

Full size image

Jill’s sessions took place in the clinic. For baseline and video modeling, Jill scored a 0 or 1 because she either touched the pill container or remained in the room with the container. She never attempted to open the containers or remove the pills to touch or ingest them during in situ assessments. In IST, Jill scored a 1, 0, and a 3 before her data stabilized at a score of 1. It was hypothesized that Jill was not engaging in the target behaviors because she was accessing attention from a researcher during in situ training. Therefore, the incentive condition was implemented for her, and she scored a 3 immediately in her first assessment following implementation of the incentive condition, but then, her data decreased to scores of 0 or 1. Another priming session was conducted, and Jill scored a 3 for the following three in situ assessments. Jill did not maintain the skills in her 1-week follow-up assessment, so a priming session was done until Jill achieved a score of 3 in the following assessment. Jill maintained the skills in her 3 and 5-week follow-up assessments.

Carl’s sessions took place in both the home and clinic throughout all phases of the study. In baseline, Carl scored three consecutive scores of 1 in the clinic because he stayed in the room with the pill container, but did not touch the pill container. In the following baseline assessments that took place in the home, Carl touched the pill container, scoring a 0 for both assessments. Carl never attempted to open the container to remove, touch, or ingest the pills. Video modeling for Carl was not effective; he touched the pill container in all three assessments after watching the videos. One IST session was implemented after the third video modeling assessment, and Carl mastered the skills for the final three consecutive assessments. He maintained a score of 3 at his 1, 3, and 5-week follow-up assessments.

Nicole’s sessions also took place in both the home and clinic throughout all phases of the study. Her first baseline assessment took place in the clinic, and she scored a 0 because she touched the pill container, opened the container, removed the pills from the container, and lined the pills up on the table before putting the pills back into the container and closing the lid. She never ingested the pills. Her scores then increased to two consecutive scores of 1 and then a score of 2 for her assessments before decreasing to three consecutive scores of 0 in both home and clinic settings. Nicole scored a 1, 2, and 0 for her assessments in the video modeling condition. IST was implemented during the third video modeling assessment in which Nicole failed to execute the safety skills. Nicole required multiple in situ trainings before mastering the skills, but she maintained the skills in her 1, 3, and 5-week follow-up assessments.

Table 1 shows the data collected on the number of prompts for participants to attend to the videos, the number of incorrect answers to the narrator’s questions, and number of verbal prompts to evoke correct responses during the videos. The data show that the number of prompts to attend to the videos, the number of prompts to evoke the correct answer, and the number of errors decreased from the viewings of video 1 to video 3. While viewing video 3, Carl and Nicole had no errors and required no prompts while Jill had just one error and required just one prompt.

Table 1 Responses of participants during the video modeling intervention
Full size table

Parents completed the side effects and social validity questionnaire after their child mastered the poison prevention skills. All parents strongly disagreed that their child appeared to touch medications more after participating in the study. One parent reported that her child was overly communicative about medications after participating in the study; however, this parent disagreed that her child was obsessed with finding medication as a result of participating in the study. The other parents reported that their child did not appear to talk about medications more, be obsessed with finding medication, or be scared of medications as a result of participating in this study. Parents did not recognize any other behavior changes as a result of their child participating in the study. All parents reported that their children did not report the presence of medication outside of training during or after completion of the study. This may have been due to parents keeping medications inaccessible to their children. All parents reported that they were very pleased with their child’s participation in the study, were satisfied with the way researchers communicated about what was going on in the study, and considered teaching poison prevention skills to children to be very important.

Discussion

The purpose of this study was to evaluate video modeling for teaching poison prevention skills to children diagnosed with ASD. Results showed that video modeling was not effective for any of the participants but that IST was effective for two of three participants (the incentive condition was needed for Jill). In situ assessment results were consistent with previous research indicating that children behave dangerously in the presence of safety threats (Dancho et al. 2008; Jackman et al. 2001; Poche et al. 1981). For instance, all of the children touched the pill container during assessments, and one participant opened the containers, touched the pills, and removed the pills from the container. None of the participants ingested the pills. These findings suggest that future research should continue to evaluate effective intervention methods to teach children to avoid medication because children did not exhibit safety skills in the presence of medications before intervention.

Although research indicates promising results for the effectiveness of video modeling with some skills for children with ASD (Acar and Diken 2012; Charlop-Christy et al. 2000; Godish et al. 2017), it was not effective for this study. This finding could have occurred for several reasons. First, the participants in this study were functioning at a lower level than the participants in Godish et al. (2017). Second, the threat in this study was a physical threat, whereas the threat in Godish et al. was a social threat. Perhaps children with ASD respond differently depending on the source of the safety threat.

Godish et al. (2017) showed that video modeling was effective for teaching abduction prevention skills, but their three participants were identified as having high functioning ASD and more typical verbal abilities. There is no current research to determine if video modeling is effective in teaching safety skills to individuals with lower functioning ASD. Future research should continue to evaluate whether individuals with higher functioning ASD and verbal abilities benefit more from video modeling than individuals with lower functioning ASD and limited verbal abilities. It is possible that video modeling was not effective for this study because the participants did not possess as advanced verbal repertoires as those participants in the Godish et al. study. However, Carl had an advanced verbal repertoire and demonstrated all of the skills during the informal probe assessment, but did not master the skills with video modeling.

Second, this study is consistent with Morgan and Miltenberger (2017) in demonstrating that video modeling was not effective in teaching children diagnosed with ASD to avoid a physical safety threat. Although video modeling has been shown to be effective in teaching a variety of appropriate behaviors and ways to respond to social interactions, such as abduction lures, it may not be effective in teaching children how to respond to physical items that pose a threat (Carroll-Rowan and Miltenberger 1994; Gunby et al. 2010; Poche et al. 1988). Future research should continue evaluating video modeling to teach safe responses to social and physical safety threats and see if differences in results exist.

The participants in this study required several prompts to correctly respond to the narrator’s question during the videos. In Godish et al. (2017), the participants responded correctly 98 to 100% of the time during the interactive components of the videos and all participants mastered the abduction prevention safety skills after watching the videos. In Morgan and Miltenberger (2017), two out of three participants required a high frequency of verbal prompts to respond correctly during the videos and did not master the gun avoidance skills with video modeling. The other participant answered 17 out of 18 questions correctly during the videos and mastered the gun avoidance skills with video modeling. Therefore, responding correctly and independently during the interactive components of video modeling may be an indicator of video modeling’s effectiveness for teaching safety skills. Future research should continue to assess the relationship between the frequency of prompts required during the interactive components of video modeling and mastery of the safety skills.

Teaching children to avoid pills rather than firearms or abduction lures is relatively novel in safety skill research (Gatheridge et al. 2004; Himle et al. 2004; Johnson et al. 2005; Miltenberger et al. 2004; Poche et al. 1981, 1988). Dancho et al. (2008) had three children participate in a 30-min group safety training that included BST components where the researcher instructed the children to ask permission from an adult before eating or consuming anything, modeled the skills, had the children rehearse the skills several times, and provided praise or corrective feedback for performances. The participants were also required to say the correct behavior when asked what to do if they found food, drinks, or other ambiguous substances. The children did not refrain from engaging in opening the pill or unlabeled liquid containers and ingesting the substances, however, until in situ feedback and response interruption were implemented (Dancho et al. 2008). Therefore, teaching children to avoid safety threats such as pills may require more intrusive methods such as IST that mimic natural scenarios and reduce reactivity rather than methods that include role play scenarios, stating the poison prevention behaviors, or other antecedent methods.

Although several studies have been conducted to evaluate BST, IST, and video modeling for teaching safety skills with typical children (Dancho et al. 2008; Gatheridge et al. 2004; Himle et al. 2004; Johnson et al. 2005; Miltenberger et al. 2004; Poche et al. 1981, 1988), little research has been conducted to evaluate these methods with children diagnosed with ASD (e.g., Bergstrom et al. 2014; Godish et al. 2017; Gunby et al. 2010; Gunby and Rapp 2014; Morgan and Miltenberger 2017). Furthermore, this is just the second study to evaluate behavioral procedures for teaching children to avoid pills as a safety threat rather than firearms or abduction lures (Dancho et al. 2008). Although Dancho et al. (2008) taught poison prevention skills to 15 typically developing preschool children, this is the first study to evaluate teaching children diagnosed with ASD to avoid pills. Because video modeling in this study failed, it may be a priority for future research to evaluate IST alone, BST, and other procedures that may be effective in teaching children diagnosed with ASD to avoid poison hazards.