Protective procedures in functional analysis of self-injurious behavior: An updated scoping review

Abstract

Despite the efficacy of functional analyses in identifying the function of challenging behavior, clinicians report not always using them, partly due to safety concerns. Understanding how researchers employ safeguards to mitigate risks, particularly with dangerous topographies like self-injurious behavior (SIB), is important to guide research and practice. However, the results of a scoping review of functional analyses of self-injurious behavior conducted by Weeden et al. (2010) revealed that only 19.83% of publications included protections. We extended the work of Weeden et al. to determine whether reporting has improved. We observed increases in all but two types of protections reviewed by Weeden et al. Additionally, we included new protections not reported by Weeden et al. In total, 69.52% of the studies included at least one protective procedure and 44.39% specified that the protections were used for safety. It appears that reporting has increased since Weeden et al. called for improved descriptions of participant protections.

Self-injurious behavior (SIB) includes a variety of responses directed toward oneself that have the potential to cause or have a history of causing tissue damage. It is perhaps one of the most dangerous and debilitating forms of challenging behavior exhibited by individuals with intellectual and developmental disabilities. Although prevalence estimates vary, recent reports suggest that nearly 45% of individuals with autism spectrum disorder (Steenfeldt-Kristensen et al., 2020) and 58% of individuals with severe intellectual disabilities (Davies & Oliver, 2016) exhibit SIB. Furthermore, SIB can emerge early in development (e.g., Fodstad et al., 2012), beginning as self-stimulatory stereotypy or proto-injurious behavior before developing into a topography that produces actual tissue damage (e.g., Berkson et al., 2001; Richman & Lindauer, 2005).

Prevalent topographies of SIB include head-hitting, head-banging, self-biting, self-scratching, and self-hitting (Kahng et al., 2002; Shawler et al., 2019). Self-injurious behavior can cause serious harm, ranging from transient contusions and lacerations to permanent disfigurement and even vision loss (de Winter et al., 2011; Hyman et al., 1990; Rooker et al., 2018, 2020). For example, Hyman et al. (1990) conducted a retrospective chart review of SIB exhibited by 97 children admitted to an inpatient hospital unit for the assessment and treatment of challenging behavior. At the time of the study, physical injury was documented for 77% of the cases. Soft tissue damage, such as abrasions (observed on 49.5% of the children), was the most reported injury type. Approximately 2.1% of the children experienced skeletal fractures and 7% incurred injuries to the eye, such as corneal abrasions (3.1%) and retinal detachment (1%).

The current standard of care in the assessment and treatment of SIB involves first identifying the causes of SIB through functional behavioral assessment, then using this information to guide the development of an individualized, function-based behavioral intervention. Functional analysis (Iwata et al., 1982/1994) is regarded as the most rigorous approach to functional behavior assessment of challenging behavior (Rooker et al., 2015). The functional analysis is a controlled assessment wherein antecedents and consequences hypothesized to evoke and maintain challenging behavior are manipulated in test conditions and their effects on challenging behavior are directly observed to determine its function. The causes of SIB are surmised when SIB occurs at a differentially higher level in a test condition relative to the control condition of a functional analysis where the individual has noncontingent access to preferred items and attention, instructions are not issued, and challenging behavior produces no programmed consequences (except when safety measures, such as response blocking, are necessary).

Large-scale summaries of published (e.g., Melanson & Fahmie, 2023) and clinically derived (Hagopian et al., 2013) functional analyses reveal that they are highly effective in identifying the functions of SIB. However, due to repeated exposure to potentially evocative events, there is inherent risk in conducting a functional analysis of SIB despite modifications prioritizing precision and safety. To examine risk of injury in a retrospective chart review, Kahng et al. (2015) used the Self-Injury Trauma Scale (Iwata et al., 1990) to document the frequency and severity of injuries that occurred during the functional analysis and during the same period outside of it for 99 individuals admitted to an inpatient hospital. Results indicated that although the rate of injury may be higher during the functional analysis than outside of it, the severity of injuries was generally low across both contexts. The authors noted that many safeguards were in place to protect the patients in their study, and they concluded that functional analyses of SIB can be safe when conducted using the proper precautions.

Several tactics can be employed to mitigate risks that are associated with SIB during assessment. In fact, Iwata et al. (1982/1994) outlined numerous safety precautions taken in their work with individuals with SIB. They excluded individuals whose SIB put them at imminent risk for injury based on medical exams conducted prior to initiating the study, sought approval from an ethics board, included individualized termination criteria, conducted medical examinations after terminated sessions and after every fourth session, and had medical personnel intermittently observe sessions. In addition to reporting on the safety precautions, Iwata et al. noted the types of injuries that occurred during the functional analyses as well as the number of times and for whom sessions were terminated. Following a terminated session, the medical team was consulted and either cleared the participant for continuation or recommended that sessions be temporarily paused. Furthermore, other researchers have reported using tactics such as protective equipment (Borrero et al., 2002), mechanical restraint (Silverman et al., 1984), padded session areas (Hall et al., 2018), and other protocols to emphasize safety (e.g., Betz & Fisher, 2011). For example, Betz and Fisher (2011) encouraged terminating session following injury and taking session holidays until any affected areas had healed or an individual was medically cleared to continue. Regarding particularly severe topographies of challenging behavior, Betz and Fisher also recommended the use of response blocking or protective equipment to prevent or interrupt behavior directed toward vulnerable organs (e.g., eyes) as well as conducting a no-interaction condition rather than an alone condition when feasible such that the therapist is available to block instances of SIB when necessary. Since then, several published studies have referenced the procedures outlined by Betz and Fisher when conducting functional analyses of SIB (e.g., Fisher et al., 2016; Greer et al., 2020).

Although some studies report on precautions used during the assessment process, this does not seem to be the standard. Weeden et al. (2010) conducted a scoping review of literature published between 1994 and 2008 wherein a functional analysis of SIB was conducted. The researchers examined whether the studies included individualized termination criteria and nine safety precautions: response blocking, institutional review board (IRB) approval, protective equipment, padded test areas, few and short sessions, medical examination, mother in room, parent as therapist, and sessions monitored by medical personnel. Weeden et al. found that only 7.76% of the 116 publications reviewed included notations of termination criteria and 19.83% included at least one of the other nine protections evaluated. The authors noted that this was likely an underrepresentation of the actual use of protections during functional analysis of SIB; however, they also noted the need for authors to report on this information so the use of protections can be evaluated and, when applicable, replicated.

The safety precautions that Weeden et al. (2010) described in the literature were not exhaustive, and the authors did not delineate certain general categories of protections. For example, although Weeden et al. reported on the use of protective equipment generally, they did not differentiate between types of equipment, such as mechanical restraint (e.g., arm splints), helmets, or protective clothing. Additionally, Weeden et al. did not describe whether self-restraint items were used or if consent or assent was obtained prior to participation. Finally, the only modification to the functional analysis reported to maximize safety was few and short sessions (i.e., a brief functional assessment; Northup et al., 1991). However, other modifications to the functional analysis have been proposed to increase both safety and efficiency (Poling et al., 2012). For example, conducting a precursor functional analysis (e.g., Borrero & Borrero, 2008; Fritz et al., 2013; Herscovitch et al., 2009; Smith & Churchill, 2002) in which consequences are delivered following a precursor rather than challenging behavior can help reduce risk to the individual by limiting the amount of challenging behavior during the course of the assessment. Similarly, the interview-informed synthesized contingency analysis includes an open contingency class such that lower severity behaviors are reinforced, and it includes brief and relatively few test and control sessions (e.g., Hanley et al., 2014; Jessel et al., 2019).¹ The occurrence of challenging behavior may also be limited by using a latency-based functional analysis where sessions are terminated following the first instance of challenging behavior (e.g., Thomason-Sassi et al., 2011). In a trial-based functional analysis, antecedents and consequences are presented discreetly embedded in trials that are often conducted in the natural environment (e.g., Bloom et al., 2011; Sigafoos & Saggers, 1995). Trial-based functional analyses may minimize risk by using an occurrence measure that requires only one instance of challenging behavior to terminate the trial. Each of these modifications to the functional analysis of challenging behavior are important advancements, and researchers have suggested that they may improve both efficiency and safety in assessment.²

Most of these procedural modifications to the functional analysis occurred after the publication of Weeden et al. (2010). It is possible that researchers have heeded the calls of Weeden et al. to improve reporting practices; moreover, they may have done so by including some of these modified and, arguably, safer functional analyses. Thus, the purpose of this study was twofold. First, we replicated the scoping review conducted by Weeden et al. to provide an update on the percentage of studies in which researchers report participant protections for functional analyses of SIB. Second, we extended the findings of Weeden et al. by expanding the types of protective procedures assessed throughout the course of the review and evaluating whether certain variables, such as noted injuries or topography of SIB, were associated with notations of protections.

METHOD

Study identification and inclusion criteria

We followed the guidelines described by the Preferred Reporting Items for Systematic Reviews and Meta-Analysis-Scoping Review (PRISMA-ScR; Tricco et al., 2018). Supporting Information A includes the PRISMA-ScR checklist, Supporting Information B through G includes outcomes coded in this review for each study, and Supporting Information H includes the references for all studies that met inclusion criteria.

The inclusion criteria, adapted from Weeden et al. (2010), were as follows: (a) the article was a peer-reviewed empirical study (b) that described a functional analysis of SIB for at least one participant with an intellectual or developmental disability (c) where the data were not reported to have been published previously. Thus, we excluded (a) nonempirical or non-peer-reviewed articles such as reviews or commentaries and dissertations; (b) articles that did not include a functional analysis, articles that made reference to but did not describe a functional analysis (e.g., indicated that a functional analysis or functional assessment was completed but did not provide a description of the methods or supporting citation), or articles that included only an indirect or descriptive assessment; (c) articles where the challenging behavior targeted in the functional analysis did not include SIB for at least one participant; (d) articles where no participant had an intellectual or developmental disability; or (e) articles where data for any participant who otherwise would have met inclusion criteria were reported to have been published elsewhere (data were not original to the article). We also excluded articles that were published before 2009 or after 2022 and those that were published in a language other than English. If multiple topographies targeted within the functional analysis were collapsed into one reinforced class, the study was included if SIB was a member of that reinforced class. In some cases, articles included multiple participants but only some participants met all criteria for inclusion; we extracted data from only those participants who met all inclusion criteria.

As the first step of the review, we completed two searches to identify relevant studies. The first included a literature search completed June 14, 2023, on the EBSCOhost research platform using the Scopus and PsychInfo databases. Specifically, we conducted a full-text search using the terms (“SIB” OR “self-injur*”) AND “functional analys*” between the years 2009 and 2022. This resulted in the identification of 445 studies, 93 of which were duplicates. With duplicates removed, the initial database search resulted in 352 unique studies. However, we noted that many articles that should have been identified in this search were not among the 352 obtained results, including studies by authors of this scoping review.³ Therefore, as a second step of the review, we completed a search of each journal listed in Weeden et al. (2010; along with several relatively newer behaviorally oriented journals) using the same Boolean operators, search terms, and years used in the first step. This was completed on August 29, 2023. Step 2 yielded 1,080 studies, 202 of which were duplicates. Thus, we reviewed an additional 878 studies from the second search. In total, we identified 1,230 unique studies across the two search steps.

Figure 1 includes a PRISMA flowchart for the review process. To identify as many studies with functional analyses targeting SIB as possible, studies were included in the full-text review if the abstract indicated that the study (a) targeted SIB specifically or challenging behavior more generally, without listing any particular topography, and the abstract noted that the study included (b) a functional analysis or functional behavioral assessment or (c) a behavioral intervention. That is, studies were included in the full-text review if the abstract noted generally that challenging behavior (or some variant thereof, such as “disruptive behavior” or “problem behavior”) was targeted; studies that specified that the target response was something other than SIB, such as elopement or stereotypy, were excluded. Also, we included studies in the full-text review if the abstract did not mention that a functional behavioral assessment was completed but did note that a behavioral intervention was conducted because we hypothesized that the study authors likely reported on an assessment to inform the development of the intervention. Figure 1 denotes the number of studies excluded and reasons for exclusion. For the 10 studies that were published outside of 2009–2022, six were published before 2009 and the date was missed during the abstract search (simply due to human error); the correct date was noted during the full-text review, and the articles were thus excluded. The remaining four were published online during 2022 but in print during 2023 and were thus excluded. After review, 187 articles met inclusion criteria.

FIGURE 1.

PRISMA flowchart. PRIMSA = Preferred Reporting Items for Systematic Reviews and Meta-Analyses; FA = functional analysis; SIB = self-injurious behavior. IDD = intellectual or developmental disability.

Coding of participant data and data extraction

A coding system was developed to synthesize data from various aspects of the studies. Generally, data were extracted for (a) study characteristics; (b) safety procedure characteristics; and (c) descriptive characteristics of the participants, settings, and functional analysis outcomes. The coding system was developed prior to initiating the search and was then modified if additional protections were identified during the coding process that did not fit one of the previously established categories. If we identified a protection not in our initial coding system, we added it and recoded previously coded studies. Because the focus of our review was on SIB, we coded only participant, and not staff, protections.

Study characteristics

For each study, data collectors recorded the journal title, year of publication, and whether the study was prospective or retrospective. We differentiated between prospective and retrospective articles because we hypothesized that authors may be less likely to report safety procedures for retrospectively analyzed data, perhaps because the functional analyses were completed by other clinicians or they had minimal information about the safety procedures used. We also hypothesized that researchers may be more likely to report on safety procedures when SIB was the primary subject matter of the article than when SIB was not the focus. For example, SIB was considered the primary subject matter for articles specifically examining functional analysis outcomes for SIB (e.g., Hall et al., 2014, examined functional analysis outcomes for skin picking by individuals with Prader–Willi syndrome) or for studies examining subtypes of automatically maintained SIB (e.g., Hagopian et al., 2015). We also characterized SIB as the primary subject matter when researchers examined a particular type of treatment explicitly designed for use with SIB (e.g., Davis et al., 2013, examined the use of noncontingent matched stimuli in the treatment of automatically maintained SIB). Self-injury was classified as a secondary subject matter for articles in which SIB was just one of multiple topographies of challenging behavior for which assessment and treatment was conducted, such as a study comparing outcomes of different types of functional analyses without regard to the topographies of challenging behavior (e.g., LaRue et al., 2010, compared outcomes from traditional and trial-based functional analysis outcomes, and participant topography of challenging behavior was not essential to the purpose of the study). Thus, we coded a “yes” that SIB was the primary subject matter if SIB or some variant thereof (e.g., a specific self-injurious topography was noted, such as hand-mouthing or head-banging) was listed in the article title, study purpose statement, or keywords of the article. We coded “no” if SIB was not listed in any of those places.

Safety procedure characteristics

As previously noted, Weeden et al. (2010) evaluated the following protections: specific termination criteria, response blocking, IRB approval, protective equipment, padded test area, few and short sessions (brief functional analysis), medical exam, mother in the room, parent as therapist, and sessions monitored by medical personnel. It is possible that some of these procedures may not enhance safety or could even increase risk (e.g., caregiver as therapist or in the room). However, one goal was to replicate the procedures of Weeden et al.; therefore, we included the same protections as described in Weeden et al. even if the protections were perhaps not distinctly used to enhance safety. We extended the safety procedures reported by Weeden et al. by including additional types of participant protections (e.g., different variations of functional analyses, participant assent). Furthermore, we categorized all protections as follows: environmental safety protocols, protective equipment, mechanical restraint, staffing protocols, session parameters, and ethical protections. Table 1 includes a description and example of each category. Within each category, we specified the type of protections noted.

TABLE 1.

Protective procedure categories and examples.

Category	Description	Examples
Environmental safety	Modifications made to the physical environment, to monitoring of sessions, or as a response to SIB to increase safety	Padding within the environment (walls, floors, tables), use of mats, medical intervention (exams or monitoring), response blocking
Protective equipment	Protective equipment: Items that do not restrict movement and are worn by the participant to decrease the risk of injury caused by completion of the self-injurious response	Helmets, arm or shin guards, padded gloves, protective dressings to cover wounds
	Self-restraint: Permitted self-restraint or provided self-restraint items that could be worn by the participant that were reported to be associated with reduced levels of self-injury	Items could include (but were not limited to) sweaters wrapped around arms, airplane pillows
Mechanical restraint	Items that restrict movement to prevent the occurrence of a self-injurious response	Arm restraints (with varying levels of flexion), Posey mitts, wrist guards
Staffing protocols	Procedures specific to therapists in session that could be associated with increased safety	Training in functional analysis procedures, caregiver observing session, caregiver-conducted sessions, increased therapist proximity, additional staff in the vicinity
Session parameters	Procedures specific to conducting the functional analysis that could be associated with increased safety; modifications to the functional analysis methodology that minimize occurrence of self-injury	Brief-FA, latency-FA, precursor-FA, trial-based FA, IISCA/PFA, termination criteria, brief session durations (less than 5 min or if otherwise noted for safety)
Ethical protections	Procedures related to protections of human subjects in biomedical and behavioral research	Approval by an ethics board or institutional review board, notations that the study was in compliance with ethical standards of research, caregiver consent, participant assent

Note: SIB = self-injurious behavior; FA = functional analysis; IISCA/PFA = interview-informed synthesized contingency analysis/practical functional assessment.

Environmental safety included modifications to the physical environment, monitoring of sessions, or response to SIB in pursuit of safety (e.g., padded session area, sessions monitored by medical personnel, response blocking of SIB). Protective equipment included two types of protections: protective gear worn by the participant and access to self-restraint or self-restraint items. Protective gear included items worn by the participant to protect the participant from injury but that did not restrict the participant’s movement (e.g., protective dressings to cover wounds, helmets, arm sleeves). The specific type of protective gear was also noted when possible. Some individuals who engage in SIB may also engage in self-restraint, which includes responses such as sitting on one’s own hands or wrapping one’s hands in clothing (Hagopian et al., 2015). Thus, we also wanted to review whether self-restraint items were provided or self-restraint was permitted within session, as these items may decrease the overall level of, and therefore risk inflicted by, SIB. Self-restraint items were those that the individual could voluntarily wear or place their arms or hands in or around that were reported to be associated with reduced levels of SIB. Permitting self-restraint or providing access to self-restraint items was included in the protective equipment category because (a) they served a protective function and (b) were voluntary responses (i.e., the participant could choose to engage in self-restraint or use self-restraint items). Mechanical restraint included items worn by the participant to protect them from injury but differed from protective equipment in that (a) the items restricted movement such that the individual could not complete the SIB response to cause harm and (b) they were applied on the participant (e.g., arm restraints) rather than the participant electing to engage with them. Staffing protocols included procedures related to the therapists in session that could increase the safety of the functional analysis such as having the caregiver present in the room during the functional analysis or having them conduct the session, increasing therapist proximity to the participant, or increasing the therapist-to-participant ratio (e.g., two therapists to increase the feasibility of response blocking). Session parameters included modifying the functional analysis methodology as a means of reducing risk during the functional analysis (i.e., brief, latency, trial-based, synthesized, or precursor functional analysis), using specific termination criteria, or conducting an abbreviated analysis or using shorter session durations. Functional analyses routinely range in duration. For example, in a review of functional analysis research conducted by Melanson and Fahmie (2023), 55.5% of studies included a maximum session duration of 5 min. Furthermore, of the 17.2% of studies that included a session duration other than 5, 10, or 15 min, the vast majority (79.7%) reported less than 5 min. These outcomes indicated that more than half of all functional analyses summarized by Melanson and Fahmie used short session durations, 5 min or less. Finally, ethical protections included obtaining approval from an ethics board or IRB, obtaining caregiver consent, or obtaining participant assent. Like Weeden et al. (2010), we also noted whether the authors reported having to terminate sessions.

For each study, we reviewed the text and coded the presence of the safety procedures outlined above regardless of whether the authors noted that the procedure was in place specifically for safety. That is, it was possible that a potential safety procedure was present in some articles but the authors did not explicitly acknowledge that it was included as a means of reducing risk. For example, response blocking was included in many studies identified in the current review but the study authors did not always note that they included it to protect the participant from injury. In the one exception to this rule, “short session duration” was only coded if the authors noted that the shorter durations were for safety or to reduce risk. We made this exception because so many studies include short session durations (see above) and we did not want to artificially inflate the outcomes of the current review by reporting every study that used a session duration of less than 10 min as a participant protection.

For the data analysis, we evaluated the general prevalence of procedures commonly used for participant protection as well as the prevalence of safety procedures that were specifically described as such. Therefore, for each study we coded whether it provided (a) specific or (b) general safety information. Specific safety was coded if the authors explicitly noted that at least one of the procedures was implemented for safety (or using similar terms related to risk reduction or protection). General safety was coded if any potential safety procedure was included based on the categories in Table 1 but the authors did not note that it was a specific safety procedure (see exception for the few and short sessions protection). Any reference to obtaining IRB or ethics board approval or conducting the study in accordance with the 1964 Helsinki Declaration (World Medical Association, 2013) and its later amendments anywhere in the body of the article or backmatter was categorized as “specific safety” because the guiding principles of ethics boards and the Helsinki Declaration include promoting the health, well-being, and rights of the patient or research participant. Safety is inextricably linked to these processes, as described in the Belmont Report (National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research, 1979) and the Helsinki Declaration (World Medical Association, 2013).

Participant, setting, and functional analysis characteristics

For each participant included in the study, data collectors recorded demographic information related to the participant including the individual’s gender, age, race, ethnicity, and genetic or psychiatric diagnoses. We also recorded the presence or absence of autism spectrum disorder and presence or absence of intellectual disability for each participant. For participants for whom an intellectual disability was described, we noted the level reported or “level not reported” if an intellectual disability was described but the level was not noted. In many cases, an intellectual disability or level was not explicitly noted but the researchers reported that the participant had limited to no communication skills or performed daily living skills at a level that was below expectations given their age. We classified these cases as having “reported deficits in communication or activities of daily living.”

Data collectors recorded the topography of SIB targeted in the functional analysis and whether it was noted to have produced any injuries prior to or during the functional analysis. Topographies of SIB were categorized in a manner similar to that in prior research on SIB (e.g., Hagopian, Falligant, et al., 2023) and included head-directed SIB, body-directed SIB, mouth-directed SIB, and skin-directed SIB. We also included ingestion-related SIB because a few topographies of SIB did not fit within the other four categories. Head-directed SIB included any descriptions of the participant using any part of their body or an object to hit their head or poke their eyes (it excluded picking, pinching the head or face, and mouth-directed behavior). Body-directed SIB included descriptions of the participant using their hands, legs, mouth, or objects to hit or bite parts of their body (it excluded pinching, scratching, or behavior directed to the head and mouth). Mouth-directed SIB included descriptions of the participant using their hands or objects to hit, pinch, scratch, or pick at any portion of the mouth, including hand-mouthing. Skin-directed SIB included descriptions of the participant using their hands or objects to pinch or scratch the skin (it excluded self-biting, which was categorized as body-directed SIB). Additionally, we coded ingestion-related SIB for any descriptions of the participant inserting items into their nose, mouth, or ears or ingesting items. For injuries, we recorded data separately on whether the authors noted an injury prior to the functional analysis (by history) or during the functional analysis. Data collectors recorded “yes” if the authors explicitly noted that an injury had been reported or occurred, “no” if the authors explicitly noted that no injury had been reported or occurred, or “unknown” if the authors made no notation regarding injury.

Finally, data collectors recorded the setting where the functional analysis was conducted and the functional analysis results. Settings included the participant’s home or residential placement, an outpatient or university-affiliated center or clinic, a school, or an inpatient hospital. If multiple settings were used in the same study, this was reported as “multiple.” Functional analysis results were recorded based on the authors’ reports; we did not independently interpret functional analyses. Socially mediated functional analysis outcomes were coded as “attention” (which included attention or divided attention), “tangible” (which included food or toys), “escape,” “synthesized” (which included two or more social tests in one condition), or “idiosyncratic” (e.g., mands function; see Schlichenmeyer et al., 2013, for a review of the literature). When SIB was described as automatically maintained, we categorized it as “automatic Subtype 1,” “automatic Subtype 2,” or “automatic Subtype 3” (e.g., Hagopian et al., 2015) if the subtype was reported by the authors; if the subtype was not reported, we categorized it as “automatic.”

In some articles, the data were aggregated such that an individual’s demographic, setting, or functional analysis information could not be clearly determined. We retained those articles because they met inclusion criteria, and individual participant data for those cases was categorized as “unknown.”

Intercoder agreement

Throughout the screening process, a secondary coder assessed studies for inclusion or exclusion during both the abstract and full-text review steps. Intercoder agreement was calculated by dividing the total number of agreements by the total sum of agreements and disagreements and converting the quotient to a percentage. For the abstract review, agreement was calculated for 352 studies (28.62%); overall agreement equaled 92.90%. Agreement was calculated for 116 of the 362 studies (32.04%) that met criteria for the full-text review. Overall agreement for the full-text review was 91.38%. When disagreements occurred, the primary author met with either the second or third author to reread the abstract or text and discussed the criteria until an agreement was obtained. All conflicts were resolved following discussion.

After the final list of studies was identified, a primary and secondary coder extracted data for 40 (of the 187) studies (21.39%) that met inclusion. Coding for some variables was restricted to a list of specific response options (e.g., age was restricted to specific age ranges) or “yes” or “no” (e.g., questions such as “Was a general protection included?”). Agreements in these cases included both coders selecting the same response option; disagreements included coders selecting different response options. Coding for other variables allowed for more than one response option. For example, participant diagnoses were open-ended it to allow data coders to include all reported psychiatric and genetic diagnoses; coders could also select from multiple protections within each participant protection category. For these variables, a proportional-agreement method was used to calculate agreement. For example, if the primary coder denoted the diagnoses of both autism spectrum disorder and cerebral palsy for a participant, there would be two opportunities for the secondary coder to match. Thus, agreement for diagnosis for that individual would be 100% (2/2) if the secondary coder listed both diagnoses but only 50% (1/2) if they identified only one diagnosis. Study characteristics and safety procedures were extracted at the study level (safety procedures were often not reported for individual participants and thus had to be extracted at the study level); participant demographics and functional analysis results were extracted at the individual level for each participant that met inclusion criteria. On average, 49.38 opportunities for agreement (range: 30–264) occurred for each study. Overall agreement was calculated by dividing the total sum of agreements across studies by the total sum of agreements and disagreements across studies and multiplying by 100 to obtain a percentage. Intercoder agreement during full-text data extraction was 95.70%. Following disagreements, the second and fourth authors met to resolve any conflicts by rereading the full text and discussing until an agreement was reached. Then, if necessary, the first author resolved any additional conflicts that may have persisted following this step. Using these procedures, a consensus was reached for all disagreements.

Data analysis

Participant characteristics were coded at the participant level and included sex, age, diagnoses (including level of intellectual disability when applicable), race, ethnicity, topography of SIB, and function of SIB. We collected information about functional analysis setting(s) at the study level because not all studies reported settings in a manner that would allow us to determine settings for each participant. All study and specific protection characteristics are described at the study level. We calculated the frequency of each coded item using Excel Pivot Tables and then calculated the percentages for all items. For items reported at the participant level, we divided the total number of response options within that variable (such as a given function) by the number of participants and multiplied the quotient by 100. For items reported at the study level, we divided the number of studies with a given response option by the total number of studies (N = 187) and multiplied the quotient by 100.

Because we included the same nine specific protections and termination criteria evaluated by Weeden et al. (2010), we compared our results with theirs on these variables. Weeden et al. reported the number of studies that included the specific protections and termination criteria; we converted these values to percentages and reported on the percentage of studies that included specific protection and termination criteria. We similarly calculated the number and percentage of studies in our review that included the same nine protections and termination criteria. We also expanded Weeden et al. by including additional protections and classified all of them (i.e., the nine from Weeden et al., the 26 new protections included in this review, and the notation of following the protocol outlined by Betz & Fisher, 2011) across six categories. Thus, we also report on the number and percentage of studies that included protections from each of the six categories.

Secondary data analysis

We were interested in identifying whether certain categorical variables were more likely to be reported when safety procedures were also reported. As previously noted, we hypothesized that authors may be more likely to report protections when SIB was the focus of the study and for prospective studies. We also hypothesized that authors may be more likely to report on protections if a prior injury was noted, if the participant targeted their head (given the high risk associated with head-directed injuries), and if SIB was automatically maintained, as prior research has demonstrated that SIB maintained by automatic reinforcement is associated with a higher number and more severe injuries than socially maintained SIB (Rooker et al., 2018, 2020). Thus, we conducted a secondary analysis to examine the percentage of studies that reported on safety procedures for these variables. We supplemented the secondary analyses with chi-square tests of independence to determine whether any of these variables was associated with an increase in reporting safety procedures (see Franke et al., 2012) at a statistically significant level.

RESULTS

As noted, we were interested in identifying whether studies generally included protective procedures and whether the authors of those studies specifically designated safety as the purpose of the procedures. Figure 2 depicts the percentage of studies across years that reported on a protection and that specified it as such. Of the 187 studies that met inclusion criteria, 130 (69.52%) included at least one safety procedure. Furthermore, 83 (44.39%) specifically reported it as a protective procedure. There appears to be a slight level change in the percentage of studies with reported protections after 2013. Specifically, the mean percentage of studies for which a protection was reported from 2009–2013 was fairly stable and equaled 61.73% (range: 50%–70%); after 2014, the mean percentage of studies where a protection was reported increased to 78.18% (range: 66.67%–94.12%). When considering the 83 studies where the authors specifically reported protective procedures, we observed a fairly steady increasing trend in the percentage of studies that included a protection and specified it as such across the 14 years examined in this study.

FIGURE 2.

Percentage of studies reporting protections across years. The specific protections are reported from the 130 total studies that included at least one protection.

Study characteristics

Table 2 depicts the number and percentage of studies that met inclusion criteria, had general protections, and had specific protections across journals. Five of the 26 journals included 10 or more studies: Journal of Applied Behavior Analysis (n = 83; 44.39% of all studies), Behavioral Interventions (n = 15; 8.02%), Research in Developmental Disabilities (n = 14; 7.49%), Journal of Developmental and Physical Disabilities (n = 12; 6.42%), and Behavior Modification (n = 10; 5.35%). The remaining 21 journals included fewer than 10 studies, seven of which included only one study. The Journal of Applied Behavior Analysis had the largest number and greatest percentage of studies with general protections (n = 22; 46.81%) and specific protections (n = 30; 36.14%).

TABLE 2.

Number and percentage of studies, total protections, general protections, and specific protections.

Journal title	Studies		Total protections		General protections		Specific protections
	n	%	n	%	n	%	n	%
Journal of Applied Behavior Analysis	83	44.39	52	40.00	22	46.81	30	36.14
Behavioral Interventions	15	8.02	11	8.46	3	6.38	8	9.64
Research in Developmental Disabilities	14	7.49	13	10.00	5	10.64	8	9.64
Journal of Developmental and Physical Disabilities	12	6.42	9	6.92	3	6.38	6	7.23
Behavior Modification	10	5.35	4	3.08	1	2.13	3	3.61
Behavior Analysis in Practice	7	3.74	4	3.08	1	2.13	3	3.61
Research in Autism Spectrum Disorders	6	3.21	6	4.62	1	2.13	5	6.02
Journal of Autism and Developmental Disorders	6	3.21	5	3.85	2	4.26	3	3.61
Developmental Neurorehabilitation	5	2.67	4	3.08	1	2.13	3	3.61
Advances in Neurodevelopmental Disorders	3	1.60	3	2.31	-	-	3	3.61
Journal of Intellectual Disability Research	3	1.60	3	2.31	-	-	3	3.61
American Journal on Intellectual and Developmental Disabilities	2	1.07	2	1.54	1	2.13	1	1.20
American Association on Intellectual and Developmental Disabilities	2	1.07	1	0.77	-	-	1	1.20
Current Developmental Disorders Reports	2	1.07	1	0.77	-	-	1	1.20
Education & Treatment of Children	2	1.07	2	1.54	2	4.26		-
International Journal of Developmental Disabilities	2	1.07	2	1.54	1	2.13	1	1.20
Journal of the Experimental Analysis of Behavior	2	1.07	1	0.77	1	2.13		-
Journal of Behavioral Education	2	1.07	2	1.54	1	2.13	1	1.20
Journal of Mental Health Research in Intellectual Disabilities	2	1.07	1	0.77	-	-	1	1.20
Behavioral Development	1	0.53	-	-	-	-	-	-
Brain Injury	1	0.53	1	0.77	-	-	1	1.20
Clinical Case Studies	1	0.53	1	0.77	1	2.13	-	-
Education and Training in Developmental Disabilities	1	0.53	-	-	-	-	-	-
Journal of Applied Research in Intellectual Disabilities	1	0.53	-	-	-	-	-	-
Psychological Reports: Disability and Trauma	1	0.53	1	0.77	-	-	1	1.20
The Journal of Speech and Language Pathology and Applied Behavior Analysis	1	0.53	1	0.77	1	2.13	-	-
Total	187		130		47		83

Table 3 depicts study information. Self-injury was the subject matter of the study for 46 (24.60%) studies. Most studies (n = 166; 88.77%) were prospective. Functional analyses were most commonly reported to have been completed in an outpatient clinic (n = 45; 24.06%), a school (n = 33; 17.65%), an inpatient hospital (n = 28; 14.97%), a university clinic or classroom (n = 24; 12.83%), or across multiple settings (n = 19; 10.16%). Regarding injuries, authors of 39 studies (20.86%) noted that injuries occurred prior to the functional analysis. An injury during the functional analysis was reported in only one study (0.53%); it was noted in two other studies (1.07%) that injuries occurred during the assessment period but not during the functional analysis. Authors of one study (0.53%) reported that they terminated a functional analysis session. Additionally, authors of only four studies (2.14%) explicitly reported that no injuries occurred during the functional analysis and authors of 10 studies (5.35%) explicitly noted that no functional analysis sessions had to be terminated. No other details on injuries were provided in the studies.

TABLE 3.

Study characteristics.

Study information	Response option	n	%
Was SIB in the study title or keywords?	Yes	46	24.60
	No	141	75.40
Research type	Prospective	166	88.77
	Retrospective	21	11.23
FA setting	Outpatient clinic	45	24.06
	School	33	17.65
	Inpatient hospital	28	14.97
	University clinic/classroom	24	12.83
	Participant’s home	16	8.56
	Residential program	7	3.74
	Multiple settings	19	10.16
	Unknown	15	8.02
Injuries reported before FA	Yes	39	20.86
	Not reported	148	79.14
Injuries reported during FA	Yes	1	0.53
	Yes, outside of FA sessions	2	1.07
	No	4	2.14
	Not reported	180	96.26
FA sessions terminated	Yes	1	0.53
	No	10	5.35
	Not reported	176	94.12

Note: SIB = self-injurious behavior; FA = functional analysis.

Participant and functional analysis characteristics

Table 4 depicts participant demographics, SIB topography, and functional analysis outcomes for 1,070 participants. When individual information could be extracted, most participants were identified as male (n = 534; 49.91% of participants), primarily ages 3–12 (n = 332; 31.01%) and 13–17 (n = 136; 12.71%). Race and ethnicity were reported for only 38 (3.55%) and 7 (0.65%) participants across studies, respectively. Additionally, 735 (68.69%) participants had an intellectual disability or a reported deficit in communication or activities of daily living, without a specified intellectual disability, and 539 (50.37%) had autism spectrum disorder. We also collected information on topography of SIB. Topography was unknown for 603 (56.36%) participants. When it was reported, participants could have engaged in more than one topography of SIB; thus, the total number of topographies reported for the 467 participants with a known topography equaled 729. Head-directed SIB was the most commonly reported (n = 288; 26.92% of participants), followed by body-directed SIB (n = 235; 21.96%), then skin-directed SIB (n = 120; 11.21%).

TABLE 4.

Participant demographic information.

Demographic information	Response option	n (of 1,070)	%
Sex	Male	534	49.91
	Female	191	17.85
	Not reported	345	32.24
Age	<3	17	1.59
	3 to 12	332	31.03
	13 to 17	136	12.71
	>17	108	10.09
	Not reported	477	44.58
Race	White	21	1.96
	Black	14	1.31
	Asian/Pacific Islander	3	0.28
	Not reported	1,032	96.45
Ethnicity	Hispanic	5	0.47
	Not Hispanic	2	0.19
	Not reported	1,063	99.35
Diagnosis	ID (includes reported deficits in communication/ADLs)	735	68.69
	ASD	539	50.37
	Other behavioral	265	24.77
	Genetic/medical	164	15.33
	Other communication	8	0.75
	Vision/hearing	5	0.47
	Unknown	194	18.13
Level of ID	Severe	100	9.35
	Moderate	77	7.20
	Mild	61	5.70
	Profound	43	4.02
	Level not reported	276	25.79
	Nonspecific-Reported deficits in communication/ADLs	178	16.64
SIB topography	Head-directed	288	26.92
	Body-directed	235	21.96
	Skin-directed	120	11.21
	Mouth-directed	83	7.76
	Ingestion-related	3	0.28
	Unknown	603	56.36
Function of SIB	Automatic	273	25.51
	Subtype 1	19	1.78
	Subtype 2	32	2.99
	Subtype 3	23	2.15
	Automatic (no subtype)	199	18.60
	Escape	218	20.37
	Tangible	198	18.50
	Attention	118	11.03
	Synthesized	56	5.23
	Social (unspecified)	45	4.21
	Inconclusive/undifferentiated	37	3.46
	Idiosyncratic	10	0.93
	Unknown	292	27.29

Note: ID = intellectual disability; ASD = autism spectrum disorder; ADLs = activities of daily living; SIB = self-injurious behavior.

Table 4 also includes information on the function(s) of SIB. The results for 292 individuals (27.29%) could not be discerned because only group data were reported. Thus, we report on functional analysis outcomes for 778 individuals. Some participants were reported to have more than one function for their SIB; thus, data were extracted for a total of 955 functions (of the 778 individuals that were included). We identified 635 social and synthesized functions (66.49% of the 955 known functions), 273 automatic functions (28.59% of known functions), and 10 idiosyncratic functions (1.05%); the remaining 37 (3.87%) were inconclusive. When automatic reinforcement was identified, we also examined whether subtype of SIB was reported. Self-injury was classified as Subtype 1 for 19 participants, Subtype 2 for 32 participants, and Subtype 3 for 23 participants. Subtype was not reported or could not be determined for the remaining 199 participants.

Safety procedure characteristics

Weeden et al. (2010) comparison

Table 5 depicts the outcomes of a direct comparison with the protections evaluated by Weeden et al. (2010). Since Weeden et al., the percentage of studies including protections has increased for seven of the nine protections (albeit slightly for some). The largest changes were seen in reporting on the use of padded test areas, followed by caregivers functioning as therapists and obtaining IRB approval. Smaller changes were observed for few and short sessions, response blocking, protective equipment, and sessions monitored by medical personnel. Negligible changes were observed with the caregiver being in the room (perhaps because more were acting as therapists rather than mere observers) and medical exams (perhaps because many settings did not have access to such personnel). When examining the use of specific termination criteria, we also saw a modest increase in reporting. Weeden et al. noted that only 7.76% of studies reported termination criteria; in the current review, we found that 12.30% of studies reported using termination criteria. Finally, when comparing the total percentage of studies that included any of these nine types of protections, Weeden et al. found that at least one protection was noted in 19.83% of the studies they reviewed. In the current review, 54.01% of studies included at least one of these types of protections.

TABLE 5.

Number and percentage of studies that reported protections described by Weeden et al. (2010).

Type of protection	n Weeden et al.	%	n Current Study	%	% Change
	(N = 116)		(N = 187)
Padded test areas	4	3.45	41	21.93	18.48
Institutional review board approval	6	5.17	29	15.51	10.34
Parent as therapist	1	0.86	27	14.44	13.58
Response blocking	6	5.17	24	12.83	7.66
Few and short sessions	4	3.45	21	11.23	7.78
Protective equipment	4	3.45	14	7.49	4.04
Medical exam	2	1.72	6	3.21	1.49
Sessions monitored by medical personnel	1	0.86	6	3.21	2.35
Parent in room	1	0.86	1	0.53	−0.33
Total n studies that included a protection (excluding specific termination criteria)	23	19.83 a	101	54.01 b	34.18
Specific termination criteria	9	7.76	23	12.30	4.54

^aThe values in the column denoting the total data reported for Weeden et al. (2010) were derived from Table 1 in Weeden et al. and represent the number and percentage of studies that reported at least one type of protection (with the exclusion of specific termination criteria, which are reported separately).

^bThe values in the column denoting the total data for the current study were derived by calculating the total number and percentage of studies that reported at least one type of protection listed in this table (with the exclusion of specific termination criteria, which are reported separately).

Additional protections and categorical outcomes

We extended Weeden et al. (2010) by including additional protections, developing categories to organize all protections (those included by Weeden et al., 2010, and the new ones added in this review), and reporting on those categorical outcomes (for categories, see Table 1). Table 6 includes details on the number of studies that included each type of protection and the percentage of studies that included at least one protection from a given category. Studies could include more than one type of protection from one or more categories. The most common protections were session parameters (n = 69; 36.90%), environmental safety (n = 63; 33.69%), ethical protections (n = 48; 25.67%), and staffing protections (n = 38; 20.32%). Protective equipment was reported in 14 (7.49%) studies, and mechanical restraint was reported in 10 (5.35%) studies.

TABLE 6.

Number and percentage of studies that reported each type of protection.

Category	Type of protection	n	%
Session parameters		69	36.90
	FA modification	44
	Few and short sessions/Brief FA	21
	IISCA/PFA	13
	Latency FA	7
	Precursor FA	5
	Trial-based FA	3
	Session/topography modification	20
	Alone not conducted/no interaction instead of alone	12
	Special topography selection	4
	Session pausing/breaks	2
	Screened for automatic reinforcement	2
	Specific termination criteria	23
	Not reported	118	63.10
Environmental safety		63	33.69
	Padding (test area or therapist)	41
	Response blocking	24
	Medical exam/clearance	6
	Sessions monitored by medical personnel	6
	Covert observation	1
	Ingestion-resistant items	1
	Sanitary procedures	1
	Not reported	124	66.31
Ethical protections		48	25.67
	Caregiver consent	43
	Institutional review board approval	29
	Participant assent	5
	Not reported	139	74.33
Staffing protocols		38	20.32
	Parent/Teacher as therapist	27
	Staff in close proximity	4
	Specific training in safety	3
	Nonspecific – Cited Betz and Fisher (2011)	3
	Parent/Teacher in room or observing session	1
	Multiple therapists	1
	Presession pairing	1
	Reviewed cost-benefit analysis of FA with family	1
	Not Reported	149	79.68
Protective equipment		14	7.49
	Nonspecific – Cited Betz and Fisher (2011)	3
	Arm sleeve(s)	2
	Used self-restraint materials	2
	Protective dressing	1
	Permitted self-restraint	1
	Unspecified	8
	Not reported	173	92.51
Mechanical restraint		10	5.32
	Arm splints	4
	Nonspecific – Cited Betz and Fisher (2011)	4
	Posey mitts	1
	Unspecified	1
	Not reported	177	94.65

Note: Studies may have included more than one type of protection within and across categories. Thus, the bolded number and percentage denoted for each category represent the total number and percentage of studies that included at least one of the types of protections noted within that category. The number denoted for each type of protection within a category represents the number of studies that included that type of protection. FA = functional analysis; IISCA/PFA = interview-informed synthesized contingency analysis/practical functional assessment.

Of the 130 studies for which a protection was noted, 52 (40%) included only a single protection, whereas 78 (60%) included multiple protections. Furthermore, studies including more than one protection could have noted (a) multiple protections from a single category, (b) a single protection from multiple categories, or (c) multiple protections across multiple categories. Of the 78 studies with multiple protections noted, 15 (19.23%) included more than one protection within a single category, 24 (30.77%) included a single protection from multiple categories, and 39 (50%) included multiple protections across multiple categories. Thus, it was most common for studies to include a variety of protections that spanned multiple categories.

Figure 3 depicts the percentage of studies for which protections from each category were reported across the years. No discernable trends were observed in the percentage of studies with reported staffing protocols and environmental protections between 2009 and 2022. The year 2015 appears to be an inflection point for many of the other categories. Few if any studies included reports of protective equipment or mechanical restraint prior to 2015 (protective equipment: M = 1.36%, range: 0%–4.17%; mechanical restraint: M = 0%). After 2015, the percentage of studies with reports of these categories increased, although reporting was variable across the remaining years (protective equipment: M = 12.16%, range: 0%–25%; mechanical restraint: M = 10.47%, range: 0%–37.50%). Except for the year 2010, the percentage of studies reporting on session-parameter protections from 2009 to 2014 appears low and stable (M = 28.40%, range: 20%–50%). However, there is a notable upward shift in the percentage of studies reporting on session parameter protections from 2015 to 2022 (M = 46.19%, range: 35.71%–55.56%). Finally, there appears to be a steady increasing trend across the years in the percentage of studies where an ethical protection was reported.

FIGURE 3.

Percentage of studies reporting each category of protection across years.

We further analyzed what types of protections contributed to the changing levels and increasing trends observed in the session-parameter and ethical-protection categories, respectively. Figure 4 depicts outcomes for the three subcategories that comprised session parameters: functional analysis type, termination criteria, and modification to the session or topography. For the most part (outside of the years 2020 and 2021), there is little change in the percentage of studies that reported using specific termination criteria or that reported session or topography modification. On the other hand, the year 2015 appears to be an inflection point for functional analysis type where we observe an overall level change in the mean percentage of studies reporting on this type of protection (2009–2014: M = 13.08%, range: 0%–40%; 2015–2022: M = 34.51%, range: 25%–50%). The inset panel in Figure 4 depicts the percentage of studies where each specific type of functional analysis was reported; the change after 2015 appears to be attributable to increases in the use of the interview-informed synthesized contingency/practical functional assessment and latency functional analyses.

FIGURE 4.

Percentage of studies reporting session parameter protections across years. FA = functional analysis; Sess = session; IISCA/PFA = interview-informed functional analysis/practical functional assessment. Inset panel: percentage of studies reporting each type of functional analysis; the y-axes for the graphs in the inset have been adjusted to 50% to more clearly depict the data associated with lower percentages.

Figure 5 depicts the percentage of studies with reports of ethical protections, and the inset panel shows data for relevant subcategories: caregiver consent, IRB approval, and participant assent. Between 2009 and 2013, ethical protections consisted almost exclusively of caregiver consent. After 2013, an increasing trend was observed with IRB approval and caregiver consent; the similar trend observed with these two protections intuitively makes sense because IRBs require consent for prospective research. Relatively few studies reported on assent procedures between 2009 and 2022.

FIGURE 5.

Percentage of studies reporting ethical protections across years. IRB = institutional review board. Inset panel: percentage of studies reporting each type of ethical protection.

Secondary analysis findings

We also evaluated whether studies were more likely to include protections when (a) SIB was the subject matter of the study, (b) the study was prospective, (c) injuries were reported prior to the functional analysis, (d) at least one participant had head SIB, and (e) at least one participant had automatically maintained SIB. Table 7 depicts the outcomes from these secondary analyses. Overall, there were minimal differences in the percentage of studies that did and did not report a participant protection across most variables that were evaluated, with the exception of topography. Specifically, at least one participant engaged in head-directed SIB in 83.67% of the studies that included a participant protection versus 63.27% of studies where protections were not included. The chi-square test of independence determined that this difference was statistically significant, χ²(1) = 6.7, p = .00096, OR = 2.8, which indicated an association between participants engaging in head-directed SIB and reporting safety procedures.

TABLE 7.

Secondary analysis outcomes.

Variable	Studies that reported protections			Studies that did not report protections
	n studies	n with variable	%	n studies	n with variable	%	Chi-square values
SIB subject matter of study	130	31	23.85	57	15	26.32	0.7181
Prospective study	130	113	86.92	57	53	92.98	0.2271
Injury prior to functional analysis	130	29	22.31	57	10	17.54	0.4604
At least one participant head SIB	98a	82	83.67	49c	31	63.27	0.0096*
At least one participant automatic function	120b	33	27.50	54d	15	27.78	0.9697

Note: SIB = self-injurious behavior.

^aOf the 130 studies that reported protections, 32 did not provide sufficient information to determine the topography for any participant. Thus, data on topography were extracted from 98 of these studies.

^bOf the 130 studies that reported protections, 10 did not provide sufficient information to determine the function for any participant. Thus, data on function were extracted from 120 of these studies.

^cOf the 57 studies that did not report protections, eight did not provide sufficient information to determine the topography for any participant. Thus, data on topography were extracted from 49 of these studies.

^dOf the 57 studies that did not report protections, three did not provide sufficient information to determine the function for any participant. Thus, data on function were extracted from 54 of these studies.

^*p < 0.05.

DISCUSSION

Overall, the results of this review indicated that at least one safety procedure was noted in 69.52% (130/187) of studies that included a functional analysis of SIB. Furthermore, 54.01% (101/187) of studies included at least one of the nine types of participant protections that were also evaluated by Weeden et al. (2010). In comparison, Weeden et al. found that only 19.83% (23/116) of studies in their review reported participant protections. We also determined how many studies reported on procedures that met our classification as a protection and where the study authors described at least one safety procedure as such; only 83 (44.39%) studies did both.

We evaluated protections at the level of the study because, in most studies, protections were only reported generally and we could not always discern for which participant a given protection was used. Thus, we could not evaluate whether any patterns emerged with the use of specific types of protections aside from finding that protections were more commonly reported when a participant had head-directed SIB. Including information on the protective procedures used at the level of the individual could be particularly important as we seek to better understand how to select optimal protections. Protections, such as protective equipment, can sometimes suppress responding during the functional analysis (e.g., Borrero et al., 2002; Le & Smith, 2002; Moore et al., 2004), making the determination of the causes of SIB difficult. Therefore, optimal protections will likely include those that allow for the occurrence of SIB to the extent that it can be assessed and treatments can be evaluated while also minimizing risks of injury. We recommend that authors report specific protections at the participant level. For example, Hagopian, Falligant, et al. (2023) included a table (Table 6, p. 584) that denoted the types of individually applied protections used during functional analyses for participants who engaged in SIB. They also described safeguards employed generally across all participants.

Perhaps equally important to including information on individualized protections is information related to termination of sessions and injuries that may have occurred during the assessment. In the current review, few studies reported on this information; only 11 studies explicitly noted whether a session was terminated, and only seven explicitly noted whether an injury occurred during the functional analysis. Information related to individualized protections used and safety outcomes (e.g., termination of sessions, injuries) will be essential to future research directed at better assessing the degree to which protections are working. To that end, Supporting Information I includes a checklist that we developed to aid in documenting and reporting on protections; this could be included as supporting information in future studies where SIB is a target response.

Studies included in this scoping review also did not report on how protections, when used, were identified. Wiskirchen et al. (2017) noted that although the literature includes examples of protections used during functional analyses, there are no clear guidelines to aid practitioners in selecting the most appropriate procedures. Furthermore, a survey by Deochand et al. (2020) suggested that most respondents wanted a risk-assessment tool that would allow them to gauge the risk associated with conducting a functional analysis. A risk assessment may serve to highlight safety considerations and guide decision making for the selection of protections (Wiskirchen et al., 2017). This latter point is particularly important because selection of protections should be, to the extent possible, data driven; that is, one should not simply take a “more is better” stance when applying protections in assessment and treatment. Some protections may be detrimental to the assessment process, such as when protective equipment masks some functional analysis outcomes (e.g., Contrucci Kuhn & Triggs, 2009). Other protections may simply not be warranted based on the level or severity of the challenging behavior.

In response to the call for such an assessment, Deochand et al. (2020) developed a functional analysis risk-assessment decision tool, informed by the clinical and research experiences of the researchers and a panel of experts. The interactive tool allows individuals to input data related to four risk domains: clinical experience, functional analysis environment, support staff, and behavior intensity. Based on the responses, a risk level ranging from slight risk to high risk is identified and suggestions for mitigating risk are provided. To our knowledge, this is the first risk-assessment tool of this nature, and it represents an important step in further enhancing the safety of functional analyses of challenging behavior. However, this decision tool has yet to be empirically validated and it only provides guidelines related to risk mitigation. It does not provide a systematic process for selecting and then empirically validating efficacious, individualized safety procedures for individuals who engage in SIB. It is our belief that a systematic, protective-procedures assessment represents the essential next step in empirically driven decision making related to safety procedures employed when assessing and treating SIB.

After 2014, there appeared to be an increase in the percentage of studies that included protections. This corresponded with increases in reporting on session parameters and, specifically, increased use of latency functional analyses and the interview-informed synthesized contingency/practical functional assessment. Interestingly, there were several publications related to functional analyses that occurred between 2010 and 2014 that may also have contributed to some of these changes in reporting including Weeden et al. (2010), Thomason-Sassi et al. (2011), a special issue on functional analysis in the Journal of Applied Behavior Analysis in 2013, and Hanley et al. (2014). It is possible that the increase in studies reporting protections could be, in part, related to more studies using the alternative types of functional analyses discussed in these articles and special issues or as a response to Weeden et al. However, it is important to note that we cannot determine whether the trends are a function of changes in actual use of protections or changes related to reporting on protections (or both).

We also observed increases in reporting on ethical protections from the subcategories of IRB approval and caregiver consent, perhaps because journals are moving toward requiring this type of language. However, one area with less growth in reporting was participant assent. The topic of assent in behavior analysis has received increased attention in recent years (e.g., Breaux & Smith, 2023; Flowers & Dawes, 2023; Mead Jasperse et al., 2023; Morris et al., 2021, 2024). As noted in these articles, there may be several reasons why research concerning inclusion of assent is lagging. There are many different guidelines for obtaining assent (Flowers & Dawes, 2023), and evidence surrounding the use of specific assent procedures is lacking (Mead Jasperse et al., 2023), both of which may cause confusion. In addition, IRBs and federal policies waive assent for many individuals with disabilities, and researchers may fail to acknowledge this in publications (Morris et al., 2021). In terms of waving assent, individuals with disabilities may lack the requisite skills to make appropriate discriminations related to assent or to easily communicate their assent for a procedure (Flowers & Dawes, 2023; Morris et al., 2021). Despite the noted difficulties or ambiguity surrounding assent, it is possible that continued discussion of this important topic will result in more thorough reports on assent (even when waived), just as with IRB approval and caregiver consent.

The current study is not without limitations. First, we conducted a scoping review of the literature and, like Weeden et al. (2010), our intention was to provide a large and representative sample of reporting practices related to protections used in the functional analysis of SIB. To best achieve this, we employed a two-step search process because we noted that many studies that we knew should have been identified were missing after the first step where we searched the Scopus and PsychInfo databases. Despite our best efforts to identify all relevant articles, it is likely that studies that would have met inclusion criteria were missed. Second, and as Weeden et al. noted, the outcomes of this review reflect only those protections explicitly described in the studies. It is likely that more studies included protections than the results of this review suggest. In addition, it is also likely that protections, when described, were not necessarily reflective of all protections in place. Some protections may go unreported, perhaps because they are ingrained in the clinical standard of care. For example, specialized treatment facilities and providers who regularly serve individuals with severe SIB have likely developed several tactics for mitigating risks associated with SIB to allow it to be safely assessed and treated (e.g., Hagopian, Kurtz, et al., 2023). Thus, the outcomes of the current article may represent an underreporting of protections that are used to mitigate risk of injury. We again encourage authors to continue reporting on protections employed during the functional analysis, to do so at the level of the individual participant, and to make the use of those protections explicit in the text or as supplemental materials (see Supporting Information I for a template).

A third limitation is that we focused solely on SIB, protections for participants, and functional analysis. However, individuals with challenging behavior often engage in topographies other than SIB, like aggression (Fitzpatrick et al., 2016; Hong et al., 2018). Protections for non-self-injurious behaviors may be distinct from those used for SIB; for example, safely managing aggression may require training in very specific techniques (e.g., Wine & Newcomb, 2024). Relatedly, protections for staff or caregivers who participate in the assessment and treatment process have not yet, to our knowledge, been sufficiently examined. In addition, functional analysis is useful in that it can inform the development of behavioral interventions. Although it seems intuitive that risk of injury is reduced during intervention, we do not currently know whether this is true or whether risk of injury changes over the course of an evaluation. For example, injury may be more likely during baseline or generalization phases and perhaps less likely during initial treatment phases when schedules of reinforcement are relatively dense. Thus, additional research is warranted on risk of injury to staff, risk of injury during intervention, and how best to protect staff during the assessment and treatment process.

In summary, the results of this scoping review are promising. In 2010, Weeden et al. asked, “Where are the descriptions of participant protections?” (p. 299). It appears that many researchers have responded to calls to improve protective-procedure reporting practices related to SIB. Importantly, we observed improvements in reporting for almost all types of protections included in Weeden et al. (2010). Although there is still progress to be made, we are encouraged to report that protections are finding their way into the literature, and we hope the trend continues in this positive direction.

DATA AVAILABILITY STATEMENT

Supporting Information A includes the PRISMA-ScR checklist, Supporting Information B through G includes outcomes coded in this review for each study, and Supporting Information H includes the references for all studies that met inclusion criteria. Supporting Information I includes a checklist to aid in documenting and reporting on protections employed in research.