Measuring the Effects of Physical Exercise On Stereotypic Behaviors in Autism:Using Hierarchical Linear Model

Individuals with autism experience the core symptoms of delays in communication, delays in social skills, and the engagement in repetitive and restrictive patterns of behavior. Physical exercise has shown the ability to positively impact the core symptoms of autism. Specifically, research has demonstrated the ability to reduce stereotypic behaviors in individuals with autism. The current investigation used the Hierarchical Linear Model, an enhanced type of meta-analysis, to analyze data retrieved fromsingle-subject research design studies to identify possible moderators that increase the effectiveness of the physical exercise intervention on individuals with autism that display stereotypic behaviors. This research assessed three key areas: if different types of stereotypic behaviors responded differently to physical exercise, if different types of physical exercise are more effective in reducing stereotypic behaviors, and if different durations of exercise sessions are more effective in reducing stereotypic behaviors. The results of the current study demonstrated that type of physical exercise is a significant moderator and that types of stereotypic behaviors and duration of exercise sessions are not significant moderators to the effectiveness of physical exercise on reducing stereotypic behaviors in autism.


Introduction
Autism is considered a neurodevelopmental disorder. The Diagnostic and Statistical Manual 5 th Edition (2013) requires that individuals with autism demonstrate the core symptoms of persistent impairment in reciprocal social communication and social interaction, they must engage in restricted, repetitive patterns of behavior, interests, or activities [1]. It is suggested that autism is the fastest growing disability on the planet [2] and it is considered to be one of the most debilitating [3].
Autism is a very diverse and complex disorder. There is no known cure and there is no known single treatment. Individuals with autism often work with a team of professionals that consists of speech therapists, occupational therapists, developmental therapists, physical therapists, behavioral therapists, special education supports, and some may require medication management services. The focus of many behavioral therapies is to increase appropriate skills and decrease unwanted, problem behaviors. This is often accomplished by using a combination of what are referred to as comprehensive treatment models, treatment packages or treatment systems and focused interventions. Comprehensive treatment models consist of a set of practices designed to achieve a broad learning or developmental impact on the core features of autism. Focused intervention practices are designed to address a single skill or goal of a learner with autism.
The National Clearinghouse on Autism Evidence and Practice (NCAEP) and the National Standards Project (NSP) have developed criteria to identify interventions as evidence-based when working with individuals with autism. The NCAEP and the NSP recognize physical exercise as an evidence-based practice for individuals with autism. Research has shown that individuals with autism benefit physically and psychologically from physical exercise. Individuals with autism have experienced physical improvements by way of increases in muscular strength, doi: 10.7243/2054-992X-8-1 endurance, physical appearance, and decreases in percent body fat and body mass index [4][5][6]. They have also experienced increases in self-confidence [4].
There is a growing body evidence for the effectiveness of physical exercise on the core symptoms of autism. Stereotypic behaviors, a type of repetitive pattern of behavior, are present in nearly 90% of individuals with autism [11]. Lee et al. (2007) reported that behaviors, such as stereotypic behaviors, may create social stigma and reduce the number of opportunities for full societal integration. Stereotypic behaviors may warrant intervention if the behaviors are isolating an individual from his or her peers and if the behaviors are harmful to the individual with autism or others in the immediate environment [13].
Physical exercise has a long history of evidence in reducing stereotypic behaviors in individuals with autism. Watters and Watters (1980) were the first to use antecedent-based physical exercise as an intervention to decrease stereotypic behaviors in students with autism. Antecedent-based physical exercise intervention requires individuals to engage in the intervention prior to the recording of the desired behavior. The researchers studied the effects of antecedent physical exercise on stereotypic behaviors of five children with autism within the school setting. Watters and Watters (1980) demonstrated that 8-10 minutes of exercise was sufficient to decrease stereotypic behaviors in five students with autism.
The results of Watters and Watters (1980) have been replicated numerous times. Several studies have demonstrated that physical exercise can reduce stereotypic behaviors in people with autism. These studies have revealed a few common themes. First, studies consistently demonstrate that physical exercise does reduce stereotypic behaviors in individuals with autism [14][15][16][17][18][19][20][21][22][23][24][25][26]. Second, a number of studies have demonstrated that vigorous exercise is more effective than mild exercise in reducing stereotypic behaviors [15,16,19,21,27]. Finally, a number of studies have shown that physical exercise reduces stereotypical behaviors for short periods of time. Studies have shown that single bouts of physical exercise reduce stereotypical behaviors for 90-120 minutes before returning to baseline levels [21,28,29].
Various durations of exercise have been successful in decreasing stereotypic behaviors as well. Celiberti et al. (1997) demonstrated success with bouts of exercise that lasted 6 minutes per session. Burns and Ault (2009) demonstrated success in reducing stereotypic behaviors with bouts of exercise that lasted 5-8 minutes. Watters and Watters (1980) were successful in using physical exercise to decrease stereotypic behaviors with 8-10-minute acute bouts of exercise. Powers et al. (1992) and Morrison et al. (2011) found success with 10-minute bouts of exercise. Kern et al. (1982) were successful in using acute bouts of exercise that lasted 5-20 minutes in length. A number of studies used 15-minute bouts of exercise [16,27,19,30,32]. Many studies exercised participants for 20 minutes per exercise session [19,20,24] Finally, studies have described several different topographies of stereotypic behaviors ranging from common body rocking [14,16,19,20,25,26,27,28,29,32] to more bizarre eye gazing stereotypies [15,16,21,26,27]. Levinson and Reid (1993)  Meta-analysis research synthesizes information within three hierarchical levels; participant, procedure, and results [33]. When analyzing data with such hierarchies, statistical techniques that account for these hierarchies is essential. HLM is a complex form of ordinary least squares regression that accounts for data at different hierarchies [33]. HLM allowed this research team to analyze variance among outcome variables when the predictor variables were at varying hierarchical levels [33]. The current investigationutilized HLM to explore the following research questions: 1. What is the overall effect of physical exercise on stereotypic behaviors in individuals with autism?

Inclusion Criteria
The current investigation established the following inclusion criteria: (1) The research article must be written in English. (2) The research article must have been written between the years 1980 and beyond; (3) The research must include antecedent physical exercise as the independent variable; (4) The research must include stereotypic behaviors as the dependent variable; (5) The research must include individuals diagnosed with an autism spectrum disorder who are between the ages of 3 and 14 years old; (6) The study must be a single-subject design study that included a functional analysis to determine the function or functions of stereotypic behaviors; and (7) The research must include the effects on stereotypic behaviors attributed to physical exercise interventions expressed quantitatively so that effect sizes can be calculated.

Search Sources
The current investigation used searched the following sources to identify articles to include in this study: (1)

Search Terms
The current investigation used the following search terms to locate articles to include in this study: physical exercise, physical activity, stereotypic behaviors, self-stimulatory behaviors, self-stimulation, restricted and repetitive behaviors, autism, autism spectrum disorder, pervasive developmental disorder, Asperger Syndrome.

Data Extraction
This research team used the ScanIT software to extract the pre intervention and post intervention data points from the graphs of the researchers who did not respond to this researcher. Once the data points were extracted this research team calculated the mean pre intervention scores, mean post intervention scores, and their respective standard deviations using Microsoft excel. Once the mean pre and post intervention, and standard deviation scores were calculated effect sizes via the Tau-U method were calculated.

Analysis of Effect Sizes Estimates
The individual effect size estimates for each of the single subject design studies was calculated using the Tau-U nonoverlap method. According to Parker, et al. (2011), Tau-U is expansion of Tau, but includes an adjustment for baseline time trends, which take into consideration the number of observations in each phase as well as the expected direction of improvement (increase or decrease in the observation values). Tau-U is computed: Parker, et al. (2011) In addition to the single subject effect size estimates, Hierarchical Linear Model (HLM) was used to synthesize the results across the studies, in order to understand the overall estimated effect of physical exercise or activity on the stereotypic behaviors. If a significant effect size estimate is revealed for the introduction of physical exercise or activity, the results of this impact will be provide for types of physical exercise/ activity (aerobic, skill based, or aerobic and skill based), for the duration of the physical exercise/activity (1 to 4 minutes, 5 to 10 minutes), and for the type of stereotypic behavior measured (motoric, vocal, or both).
Hierarchical Linear Modeling (HLM) is used to assess the relative magnitude of these effects, since the number of sessions (data points) were not consistent across each individual student's measures [34]. Parker et al. (2011) suggests "the degree of sample size sensitivity depends on the absolute magnitude of the baseline slope, with steeper slopes leading to increased sensitivity" (p.1). The HLM will eliminate this potential bias by allowing for the appropriate adjustments based on the number of sessions on mean effect size data. HLM is one analytical approach that has been found to stable and appropriate for conducting meta-analytic investigations with single-case data [35].

Results
A total of n=33 articles specifically studied the effects of physical exercise on stereotypic behaviors. Six of the n=33 articles met all inclusion criteria into this study. The single subject design studies included n=13 participants. Ten of the participants were male and three were female. The total age range of the males and females combined was 3-14 years old. The total age range for the males was 3-14 years old. The age range for the females was 8-12 years old. The individual study results are provided in Table 1.
As indicated in Table 1, 14 of the individual activities resulted in significant effect size estimates, while 18 did not. In order to understand the overall impact of the introduction of physical exercise or activity, HLM is used to synthesize these mixed results into one overarching effect size estimate. The model used to produce an effect size estimate for all of the studies as well as three potential moderator variables is stated: Level-1 Model OUTCOME ij = β 0j + β 1j *(SESSION ij ) + β 2j *(PHASECD ij ) + r ij   Level-2 Model β 0j = γ 00 + γ 01 *(EXERCISE j ) + γ 02 *(DURATION j ) + γ 03 *(STEROTYP j ) + u 0j β 1j =γ 10 β 2j = γ 20 Mixed Model OUTCMM ij = γ 00 + γ 01 *EXERCISE j + γ 02 *DURATION j + γ 03 *OUTCMTY j + γ 10 *SESSION ij + γ 20 *PHASECD ij + u 0j + r ij The results of the estimation produced an excellent estimate of reliability for the intercept α =0.974. The full results for the models are provided in Table 2.
Based on the results presented in Table 2, the overall impact of the intervention is a moderate significant effect of -.43. This phase effect is significant. The average level of response is (outcome) [exp(45.58 + -.434387)=|4.04}| during treatment. The results also indicate that the number of sessions did not have a significant impact on the overall outcome. The average rate of change per session is -3.01 (B10). This increase is not significant because the p value is >.05 therefore the baseline is flat and suggesting that there is no change over time. Therefore, doi: 10.7243/2054-992X-8-1 the final estimate of response to exercise is stated Log(Y)= 48.58 -.4344, providing an effect of -4.04 on the motoric or vocal measure. Type of stereotypic behaviors and duration were not significant moderators. However, exercise type was (p<.001) was found to be a significant moderator. Phase by exercise type is computed as [Exp(phase + exercise level)] providing the following results for the three exercise types: Aerobic only=|5.68e24|; Skill only=|6.0e14|, and Aerobic and Skill= |16.4|. This indicates that the greatest impact is found with the combination of aerobic and skill.

Discussion
The purpose of the current study was to further the findings of Tarr et al. (2020) and Ferreira et al. (2019) by analyzing data from single-subject research design studies on the effects of physical exercise on stereotypic behaviors in individuals with autism. In particular, this study used HLM to assess the overall effect of physical exercise on stereotypic behaviors, the potential differing effects of physical exercise on different topographies of stereotypic behaviors, the potential differing effects of different types of exercise on stereotypic behaviors, and the potential differing effects of different durations of physical exercise on stereotypic behaviors. Overall, the effect of physical exercise on stereotypic behaviors in the current investigation was a moderate significant effect of d = -.43. The current investigation provides further evidence that physical exercise does have the ability to decrease stereotypic behaviors in individuals with autism.
In regards to type of stereotypic behavior, exercise type, and duration of exercise sessions, only exercise type was found to be a significant moderator. The current investigation concluded that aerobic only exercise produced =|5.68e24|. Skill-based exercises only produced =|6.0e14|. The combination of aerobic and skill-based exercises = |16.4|. The current investigation concluded that the greatest impact is found with the combination of aerobic and skill-based exercises. The moderators; type of stereotypic behavior and duration of exercise sessions were found to not be significant moderators.
The results of the current study indicate that a combination of aerobic and skill-based exercise is the most efficient type of exercise to reduce stereotypic behaviors in individuals with autism. This finding adds to previous research. Most of the research studying the effects of physical exercise on stereotypic has focused on the intensity and duration of the physical exercise and not the specific type of physical exercise [15,16,19,21,27] The effectiveness of skill-based exercises is supported in neurobiological and neuroscience research. Research has demonstrated that sports drills, skill-based, motor fitness, and/or coordination training specifically impacts the basal ganglia, parietal lobe, and cerebellum in children and adults. Interestingly, various abnormalities and dysfunction within the basal ganglia [42,43,44] and the cerebellum [42,45]    The results of the current study are congruent with research associated with the effects of physical exercise on the structures and functioning of the brain, specifically the basal ganglia. A combination of aerobic and skill-based exercise routines has a greater positive impact on the basal ganglia. It may not be sufficient to claim that physical exercise has the ability to reduce stereotypic behaviors in autism. It may require that we begin to develop specific exercise routines that target specific areas of the brain. It may be even more appropriate to develop physical exercise routines that engage various areas of the brain simultaneously.

Limitations of the Study
The major limitation of the current study is the small number of research articles that met criteria for this study. A thorough search only produced six articles that met the inclusion criteria for this study. However, the method of analyses used was developed specifically for small-sample studies with individual level analyses. And, while a small number of studies is a potential limitation, it speaks to the need for additional research in the arena of aerobic and skill-based activity on stereotypic behaviors with both individual students and groups of students. A second major limitation is inherent within all meta-analytic type research is that the researcher is limited to only the information presented by each study. The current investigation incorporated those reported aspects of each study that were most likely to be related to the overall impact based on the extant autism research.

Future Research
Continued research is needed on the effects of physical exercise on stereotypic behaviors in individuals with autism. There have been a limited number of studies conducted on the effects of physical exercise on stereotypic behaviors in autism since Watters and Watters (1980). Future research is needed to identify specific characteristics of physical exercise routines that efficiently and effectively reduce stereotypic behaviors in individuals with autism. Future research is needed to decipher if certain stereotypic behaviors originate in different structures of the brain and if specific types of exercise can be prescribed to efficiently target that particular area.

Conclusion
The current study used the HLM statistical model to gain information on the effects of physical exercise on stereotypic behaviors in individuals with autism. The results of the current study demonstrate that physical exercise is effective in decreasing stereotypic behaviors in individuals with autism. Specifically, a combination of aerobic and skill-based exercise has the greatest impact on the reduction of stereotypic behaviors. This is important for attempting to identify characteristics of physical exercise routines that effectively and efficiently decrease stereotypic behaviors. The results of the current study adds further evidence that physical exercise is an evidencebased practice for improving the core symptoms of autism. Clinicians, therapists, and teachers should consider physical exercise as an effective nonpharmacological intervention for reducing stereotypic behaviors in individuals with autism.