Above and Beyond Safety: Psychosocial and Biobehavioral Impact of Autism-Assistance Dogs on Autistic Children and Their Families


 Autism-Assistance Dogs (AADs) are highly-skilled service animals trained primarily to ensure the safety of an autistic child by preventing elopement and mitigating “meltdowns”. Whereas families with AADs attest anecdotally to the psychosocial and behavioral benefits of their dogs above and beyond safety, quantitative, empirical support for these reports is lacking. The present study investigated the effects of well-trained AADs using validated clinical, behavioral, and physiological measures. We recruited families (N=13) from the top of an accredited training dog organization’s wait-list for AADs and collected pre/post-AAD data using a within subject, repeated measures design. Our findings demonstrate that, in addition to enhancing child outcomes, the integration of well-trained AADs can impact families positively across multiple domains of health and function.

). Trained service dogs may offer the further advantage of allowing for rewarding interactions using simple language that may scaffold less scripted social interactions (Solomon, 2010).
Notably, some studies have that shown autistic children undergo a reduction in salivary cortisol levels in the presence of trained service dogs Viau et al., 2010). Moreover, assistancedogs, trained for individuals with various disabilities, may serve as social catalysts, enhancing social interactions, increasing social networks, and reducing instances of social discrimination (Becker, Rogers, & Burrows, 2017;Camp, 2001;Carlisle, 2015;Mader, Hart, & Bergin, 1989;McNicholas & Collis, 2000).
An adjacent development in the burgeoning service dog industry has been the compounding demand for autism-assistance dogs (AADs) trained speci cally to address issues of physical safety. One primary source of apprehension for families of autistic children is their child's risk of elopement, de ned as a "dependent person exposing him or herself to potential danger by leaving a supervised, safe space or the care of a responsible person" (Anderson et al., 2012). An AAD team work as a triad; in public, the child wears a specially designed belt that connects to the dog's vest while an adult handler holds the dog's leash. AADs are trained to passively resist using their body weight if their child partner attempts to bolt and the tethering system prevents children from running away. Caregiver and case study reports have highlighted that AADs can prevent elopement effectively while providing a sense of security for both parents and children (Burgoyne et al., 2014;Burrows, Adams, & Spiers, 2008). Further, AADs can also be trained to provide a modi ed form of pressure touch therapy practiced by occupational therapists that has been found to help autistic individuals reduce arousal and anxiety levels (Bestbier & Williams, 2017; Grandin, 1992;Krauss, 1987). Essentially, AADs can be trained to use their weight and warmth to not only apply physical pressure on their child (e.g., leaning, laying atop) bur to also stop promptly on command.
Another major concern for families of autistic children is the wellbeing of parents/caregivers who report experiencing higher physiological stress and parenting-related stress than parents of TD children and children with other NDDs (Estes et al., 2013;Fecteau et al., 2017). Myriad factors including child characteristics and behavioral challenges (Olson et al., 2021), as well as sociocultural and economic circumstances (e.g., access to resources, stigma associated with mental health, nancial burden of care), may impact parental outcomes negatively (Bonis, 2016; Iadarola, Perez-Ramos, Smith, & Dozier, 2019) and, by extension, child wellbeing and behavior (Rodriguez, Hartley, & Bolt, 2019). However, of import to the present study is the growing focus on unanticipated secondary effects that have been observed by families with AADs. One key study reported that the contribution of service dogs to family welfare extended beyond physical safety to behavioral and psychosocial domains; parents reported improved quality of sleep and a greater sense of independence, their children exhibited fewer negative behaviors (e.g., "meltdowns", "tantrums", "bolting"), and families experienced an increase in social acknowledgement and a decrease in embarrassment or shame in public .
While the positive, multidimensional impact of these AADs has been supported by anecdotal reports, empirical data demonstrating quanti able improvements is lacking. Evidential corroboration for the bene ts of these human-canine partnerships will further the acceptance and recognition of AADs as "service animals" as de ned by the Americans with Disabilities Act (ADA) -with all provisions for access given by the U.S. Department of Justice. Hence the overarching objective of the present study has been to investigate systematically the impact of AADs on families affected by ASD. To accomplish these goals, we used standardized measures to collect longitudinal, pre-and post-intervention psychosocial and biobehavioral data from families who received highly-trained AADS. Our study contributes quantitative data from validated assessments to address the question of whether families bene t from the integration of these assistance dogs into their daily lives across multiple domains and contexts.
One critical caveat to address is that the families participating in this research study received their AADs from an internationally recognized, Assistance Dogs International (ADI) accredited, nonpro t organization that trains assistance dogs for hearing loss, mobility challenges, seizure disorders, Type 1 Diabetes, as well as ASD in children. Families are provided with the dogs free-of-charge and the economic burden and time-investment for each certi ed handler/dog team, combined with the stringent training and placement standards enforced by the organization severely limits the number of dogs placed each year. Conclusions drawn from these AAD-teams should be considered in view of their highly-specialized training and not generalized across all animals described as "therapy" or "support" with no consistent, rati ed criteria for certi cation or registration.

Methods
Study procedures were approved by the Institutional Review Board. Families were informed that their decision to participate would have no bearing on their current or future relationships with the institution or the canine training program.

Participants.
Using non-probability, purposive sampling, we recruited families from the top of a 3-5-year-long waiting list of applications to receive an AAD through the assistance-dog training organization. To apply for an AAD, children (Age: 2-7 years when applying) must have a con rmed ASD diagnosis, live within the state, and families must be physically and nancially able to take full responsibility for the dog after certi cation. The age restriction was established to accommodate the lengthy waitlist and the fact that size must be considered if the dogs will be trained to prevent child elopement. Since the potential participant pool was limited to the families who would be receiving an AAD during our period of data collection, our only exclusion criteria beyond those of the training program were that parents/caregivers be able to provide informed consent and complete questionnaires in English.
In total, we recruited 13 families to participate in the study. Final analyses included data from 11 teams; we were unable to collect post-AAD data from one family and one team experienced a change in family circumstances and had to return their dog. Mean AAD age was 2.9 ± 0.5 years when matched with a family, 45% were females, and mean weight was 62.2 ± 7.1 pounds. With the exception of one Standard Poodle, all AADs were Labrador Retrievers, Golden Retrievers, or Labrador/Golden crosses. The designated adult dog handler was the primary parent participant; 100% were mothers, 25% were from single-parent households. Secondary parent/caregiver data were collected when possible but were insu ciently powered for further analysis. Median parental education was college degree, ranging from "some high school" to "graduate degree". Median household income was "$71,000-80,000", ranging from "$31,000-$40,000" to "$100,000+". All children had a con rmed diagnosis of ASD, 41.7% were nonverbal; co-occurring neurodevelopmental conditions and additional parent and child characteristics are reported in Table 1.
Given that we would not be able to control for heterogeneity in family characteristics and child medical history and treatment, we implemented a repeated measures design that would allow us to examine changes over time within each family. We did ask parents to report ongoing medications and therapies at each assessment and no signi cant changes in ASD-related treatments between pre/post-measures were recorded. We should also note that one common factor amongst the families who chose to remain on the 3-5-year long waitlist for an AAD is a willingness and commitment to bringing an AAD into their lives and the belief that an AAD might be bene cial. Further, families would not likely apply for an assistance-dog if their child had known sensory aversions to canines (Grandin, Fine, & Bowers, 2010) that would prevent meaningful interaction. Applicants were able to make special requests for hypoallergenic breeds but those limitations could lengthen wait-times substantially. However, cortisol also binds to human nails, another keratinized matrix, and nail cortisol shows promise as an alternative retrospective biomarker of chronic stress (Liu & Doan, 2019). Because hair collection from some of the initial participants proved to be prohibitively di cult and/or not possible due to lack of scalp hair, subsequent participants were also given the option to submit ngernail clippings (Phillips, Kraeuter, McDermott, Lupien, & Sarnyai, 2021). Participants provided the same (hair or nail) samples for their pre-and post-measures. Parents were also asked to complete a questionnaire for each hair or nail sample to capture data on hair care and medication use that may affect cortisol assay results (Doan, DeYoung, Fuller-Rowell, Liu, & Meyer, 2018;Hamel et al., 2011). Ultimately, we had to limit our analysis to a subset of participants due to di culty collecting samples, low sample weight, and presence of steroid medications that may have in ated nal concentrations.
According to current estimates provided by the assistance-dog training program, AAD-teams require approximately 8-12 weeks to complete team training and certi cation. By the time they are ready for nal training, potential AADs may have already had more than 18 months of socialization, general training, assessments, and intensive training speci c to their assistance-dog career. Once the match is made, one caregiver undergoes training to become the primary dog handler and works with trainers and the AAD without their child present. When they are ready to have the dog move into the home, trainers then work with the triad (handler-dog-child) together to build their partnerships and skills in everyday life. Thus, in order to assess the real-world impact of AADs on families, participants completed the preintervention measures after being taken off the waitlist and before receiving their dogs. Two postintervention time points were included in the original study design. However, due to research and canine training restrictions during the COVID-19 pandemic, we were unable to complete all planned data collection. Moreover, we were concerned that the considerable stress and changes in routine brought on by the pandemic combined with widespread civil unrest in our regional community would mask AAD effects. Consequently, we limited our nal data set to teams who completed pre-and post-AAD assessments either before or after Spring 2020; in-person research visits were suspended in March 2020 and we pivoted to remote/online procedures only. We report herein on data collected from families before receiving their AAD and 8-12 weeks following team certi cation.
Data Analysis. Using SPSS 25.0 (Statistical Package for Social Sciences, Version 25) we conducted repeated measures analyses of variance (ANOVA) to assess pre/post-AAD changes for all measures. Both full scale and subscale scores were included when applicable. We used raw scores rather than tscores for the CBCL and SRS-2 because, at the high end of the distribution, raw scores may be more precise than t-scores (Achenbach & Rescorla, 2001;Constantino & Gruber, 2012). Signi cance levels were set at alpha = 0.05 (two-tailed). We also examined correlations between parent and child data on change in stress and cortisol levels.
Chronic Cortisol Concentration. We collected 20-50 mg of scalp hair from the posterior vertex region and stored samples at room temperature in dry and dark conditions (Cooper et al., 2012); hair was then wetted with isopropanol, minced into 2 mm pieces, and washed four times with 0.5 mL of isopropanol at room temperature for 30 seconds to remove external contamination. For ngernail samples, clippings were collected from all ten ngers and then stored and processed using an analogous protocol. Samples were dried under a nitrogen stream and weighed. Cortisol was extracted with 1 mL of methanol overnight at 55°C, 1 mL acetone for ve minutes, and then 1 mL of methanol overnight at 55°C one more time (Slominski, Rovnaghi, & Anand, 2015). Pooled solvent fractions were removed under a nitrogen stream. 1 mL of acetone was added and evaporated under a nitrogen stream to chase off the solvents' remnants.
Samples were then dissolved in in an assay diluent, randomly distributed on different plates to avoid a batch effect, and analyzed in duplicate using Salimetrics cortisol enzyme-linked immunosorbent assay (ELISA) (Miller, Plessow, Rauh, Groschl, & Kirschbaum, 2013). If readings for a sample differed by more than 10% or if readings were too high due to high concentration, the measurements were repeated; also, 5% of samples were randomly reanalyzed to ensure reproducibility.

Results
We used within-subjects contrasts to compare measures collected before families received their AAD (T1) and after they had time to complete training and integrate the AAD into their daily lives (T2). Overall, we found signi cant, positive changes over time for parent, child, and family measures. Complete results are reported in Table 3  Finally, we asked parents to describe brie y their child's relationship with their AAD.

Discussion
The overarching objective of our study has been to investigate empirically the impact of AADs on autistic children and their families. In contrast to the extant literature in AAIs, HAI, and therapy/support dogs, the participants in this study all received AADs with > 18 months of specialized service dog training whose temperaments/talents were carefully matched to families by highly-experienced trainers. Trainers were also able select for certain characteristics (e.g., hypoallergenic breeds) and tailor nal training to meet the needs of individual families. Because we were able to enroll participants shortly before they received their dog, we were afforded the opportunity to examine measurable pre/post-AAD changes across a battery of psychosocial and biobehavioral assessments. Our data further substantiate the premise that AADs confer signi cant bene ts to their families -above and beyond their initial purpose of ensuring their child's safety.
Prior research has suggested that dogs are particularly adroit at eliciting prosocial behavior and acting as social catalysts with humans, as well as reducing physiological arousal and stress in children and adults Limitations.
While we demonstrate promising outcomes for the integration of AADs into families, our data are limited in a number of ways. First, we did not collect data from a comparison group of families who did not receive a dog during the same period of time. Given the highly multifactorial nature of each family's individual characteristics, the unpredictable length of time each family remains on the wait-list, and the inexorably limited sample size, we decided to constrain the study to a single group, repeated measures design. Including a control group from further down the wait-list would require participant families to remain on the wait-list for the duration of the study collection period and we did not wish to interfere with the standard operating procedures of the training program. In particular, we did not wish to be a factor if an AAD candidate proved to be a good match for the control-family and collecting an appropriately-timed T2 assessment would delay the process of getting the AAD team started. Moreover, families who would be unlikely to receive a dog during our collection period (i.e., bottom of the 3-5 year wait-list) would include a younger cohort of autistic children and families might choose to remove themselves from the waitlist in favor of alternative options. Second, due to the limited availability of these AADs, our sample size was quite small. However, our study was designed examine pre/post-ADD changes within subjects and we were able to demonstrate signi cant, quanti able changes from T1 to T2. Third, because we were unable to collect all data from the third time point as originally planned, we were unable to evaluate the long-term sustainability of improvements. One question that could have been addressed is whether the non-signi cant ndings on the CBCL and SRS-2 would have changed after continued AAD interaction. For example, one putative mechanism could follow reduction of physiological arousal stress and increase in feelings of physical safety with the AAD to an effect on sleep. Several of our parents reported that their children have di culty sleeping which, in turn, affected their own sleep quality. Parental sleep deprivation indubitably plays a role in parent mental health and well-being, which can then impact multiple levels of family systems and behavior (Mihaila & Hartley, 2018). However, several families noted that with the AAD's presence, their children began sleeping through the night, perhaps due to an increased sense of security or their canine's de-arousing capabilities. Perhaps some AAD effects follow a more protracted time course through indirect pathways; these may not be evident until more time has passed. Another concern is that we could not control for the myriad variables that may have contributed to changes over the study timeline. For example, we cannot rule out the impact of developmental change over the study months and families maintained their ongoing treatment and medication schedules while participating. Although our T2 data demonstrated signi cant improvements in participants relative to their T1 data, we cannot be certain that changes were not due to variables such as maturation, concurrent treatments, or unknown environmental factors. Moreover, most measures were parent-report and parent-self report assessments, which may be subject to response bias. Unfortunately, more objective, task-based measures were not feasible in our functionally diverse sample. Finally, we had trouble collecting samples for cortisol assay because several participants had very short or no scalp hair. Similar issues when collecting ngernail samples arose because some individuals bit their ngernails or kept their nails quite short. An alternative option that may offset some of these issues in future studies might be the use of toenail clippings to ascertain cortisol concentration. Nevertheless, given the heterogeneity of our participants and their families, we are reasonably con dent that receiving an AAD, the one consistent change for all families during the study collection period, was a driving factor in positive outcomes.

Conclusions
To our knowledge, the present study is the rst to examine psychosocial and biobehavioral effects of AADs trained speci cally for autistic children and their families using validated and standardized measures of stress, autism symptom severity, and child behavior, in combination with a biomarker for chronic stress. These data provide quantitative evidence that AADs can have a signi cantly positive impact on autistic children and their families across multiple domains. Yet, training assistance-dogs requires a considerable, often prohibitive, investment of resources; in addition to the innumerable hours donated by volunteers, graduating one successful team can cost in excess of $55,000 over two years (Konrad, 2009). Then, after graduation, families must assume the nancial responsibilities of canine care. Unfortunately, no health insurance policies cover any of these expenses beyond the possible application of pre-tax healthcare accounts, and most training dog organizations require that families contribute at least part of the costs themselves. Further, while U.S. federal law mandates access for service animals to public areas, including schools, the ADA also requires that the animal be under the handler's control at all times. However, because public facilities are not themselves responsible for the service animals, schools do not have to provide handlers. AADs are trained to work as part of a triad, and unless an adult dog-handler is available, the dog is thus prohibited from accompanying their child to school. In light of the demonstrated therapeutic value of AADs on child outcomes as well as family health and wellbeing, we encourage policy-makers and government institutions to consider making allowances for and providing supplemental support for families who have need of a well-trained AAD. 10-item instrument that measures degree to which situations in one's life are appraised as stressful. Items query how unpredictable, uncontrollable, and overloaded respondents find their lives.  Pre/Post-AAD mean score differences on parent self-report measures demonstrating: A. improved family experiences on the AFEQ; B. reduction of parenting stress on the APSI; C. reduction of perceived stress on PSS; and D. reduction of anxiety on the STAI.

Figure 2
Pre/Post-AAD mean score differences on parent-report measures demonstrating improvements (decrease in problem scores or reduction in challenges) on the A. CBCL; B. ASQ; and C. the SRS-2.

Figure 3
Pre/Post-AAD differences in chronic cortisol concentration levels for parents and children.