This study will examine what child variables (i.e., social attention and peer interest) predict communication gains in response to a peer-mediated intervention for 40 minimally-verbal preschool children with ASD and 80 peers. The Stay-Play-Talk with iPad intervention will be implemented for 8 weeks. Gains in communication and reciprocity with peers will be measured at post-treatment and in non-treatment settings. Another goal is to determine if variance in communication gains accounted for by the two social variables differs based on the pre-treatment measurement context.

Overview. This project will use a single group pre-post treatment design across four independent cohorts of children (n=8, 10, 11, and 11 per cohort). This design is appropriate because the purpose of this study is to examine predictors of treatment response and the intervention's efficacy has been documented in an RCT.

Research Team. Dr. Bourque (PI) was the lead investigator on a NIH funded R01, examining a peer-mediated SGD intervention for preschoolers with ASD, Co-PI of a RCT peer network intervention for elementary-age students with ASD (IES), and Co-PI of two NIH grants (R01, P01) to assess communication development of preschoolers with ASD/IDD using AAC. She has almost 20 years of established relationships with local districts recruiting participants, Dr. Boyd is PI of a NIH funded R01 grant focused on outcome measure development for children with ASD, and Co-I of an R01 focused on predictors of treatment response for children with ASD. Dr. Salley has extensive experience developing eye-tracking (ET) and bio-behavioral attention measures with infants/toddlers and is PI of a longitudinal study examining the development of early social/nonsocial attention (NIH). Dr. Fleming is Director of the Research Design and Analysis Unit in the Life Span Institute. She has participated as a quantitative methodologist on 38 NIH and USDE-supported grants. Consultants to the team bring specialized expertise: (1) Dr. Shic has contributed to the design of 50 ET paradigms and 6000 recorded eye tracking datasets, the majority at-risk for or diagnosed with ASD, and is thus well-prepared to provide guidance in ET methodology and analysis; (2) Dr. McPartland is PI of the NIH-funded Autism Biomarker Consortium for Clinical Trials, a large scale effort to evaluate the utility of ET and EEG biomarkers in ASD. He will assist in administration of ET paradigms and in data acquisition.

Sample and Recruitment. A total of N=40 children between the ages of 3-6 years will participate, with n=18 recruited in Year 1 (Cohort 1=8; Cohort 2=10) and n=22 in Year 2 (Cohort 3=11; Cohort 4=11). Participants will attend preschool with access to typically developing peers. ASD diagnosis will be confirmed by the Autism Diagnostic Observation Scale-2 and DSM-V criteria, with minimal verbal skills defined as less than 20 spontaneous words or symbols. Each child will demonstrate symbol discrimination skills by intentionally selecting preferred items in 80% of trials in two 1:1 adult sessions using a matching-to-sample procedure; children who do not meet this criteria will be excluded due to insufficient skills to benefit from SGD intervention. Recruitment will occur in local preschools and the Kansas Center for Autism Research and Training registry, which contains 3000+ families, with approximately 65% individuals with ASD. The investigators are confident they can recruit 40 children given the gradual recruitment plan in Year 1 and the teams established relationships with administrators. Two peers without disabilities from the same classroom (or school building) will be recruited for each child with ASD (n=36 peers Year 1; 44 Year 2; N=80 total). Teachers will recommend peers who meet a minimum 6 of 8 behavioral characteristics related to peer social responsiveness and fidelity in the prior RCT.

Pre-Treatment Standardized Measures. Three standardized measures will be administered at enrollment: (1) the Preschool-Language Scale-5 to describe children's expressive language and auditory comprehension skills, (2) the Early Years Battery of the Differential Ability Scales-II to assess cognitive abilities across verbal, nonverbal reasoning, and spatial clusters, and (3) the Vineland Adaptive Behavior Scales III-Parent Interview to describe communication and adaptive skills. Child language and cognitive scores will be used as indicators of global child characteristics and added as covariates to analytic models. All standardized tests and social variable measures will be administered by research staff naive to study hypotheses.

Measurement Contexts. Social attention and peer interest will be measured in three contexts at one time point, prior to the start of treatment: (1) semi-structured peer communication sample, (2) naturalistic observations, and (3) eye-tracking (ET) paradigms. The SGD will be available during contexts (1) and (2), and during breaks in context (3). The first context was used in the PIs R01 to measure communication and joint attention between children with ASD and peers; thus, the administration manual and reliability procedures already exist. In this study, codes will be modified to collect social attention and peer interest data. The PI has extensive experience developing direct observation coding systems and training staff to be reliable in collecting multifaceted communication and social behavior data for this population. Along with Dr. Salley's ET experience, our team is confident that collecting reliable data across the three planned contexts is feasible.

1. Semi-Structured Peer Communication Sample: This will be administered at school, in a quiet room with the focal child and one peer; total assessment time is 25-35 min for 12 scripted activities (2-3min/task). The PI will train the examiner to administer the assessment to fidelity (i.e., 80% of steps, see Appendix A) across three practice sessions. The 12 scripts create opportunities for the child to communicate following peer presentation of each task (e.g., give bubbles with lid glued on). Coding of the two social behaviors will occur within a 1-minute start-stop window immediately after peer presentation of each task (total coding time 12 minutes).
2. Naturalistic Observations: Three preschool contexts will be observed for 10 min each over one day, for a total of 30 min. We include multiple observations during naturalistic settings that provide different communication opportunities. To provide consistency across classrooms and opportunities for the child to engage with adults and/or peers, the three contexts will be (1) snack with an adult present, (2) free play that includes only children, and (3) a small group learning activity with both peers and adults.
3. Eye-tracking Paradigms/Method: Eye tracking measures will be collected at Dr. Salley's Child Development Lab using the Tobii TX300 which has been used extensively with young children with ASD because it is non-invasive (no equipment attached to child) and robust to head movements. This desktop-mounted system records eye movements while subjects view scenes on a 17 inch monitor. Children will participate in three well-established eye-tracking tasks shown to be sensitive to individual differences and used successfully with children with ASD of varying severity levels. The Static Visual Exploration Task consists of 12 picture arrays of 12 people (adults; children) and/or objects (half are objects and smiling faces; half are object only), presented for 10 sec each in fixed order (approx. 5 min total). The Dynamic Visual Exploration Task consists of 12 silent video arrays of objects and people (adults), with each array including four videos playing concurrently in each quadrant of the screen, with face and object locations pseudorandomized (approx. 4 min). The study team will adapt this task to incorporate videos of child emotions faces, in addition to the existing adult emotion faces. The Interactive Visual Exploration Task consists of 22 silent naturalistic videos of pairs of children engaging in cooperative and parallel play with objects (including episodes of social communication) in various settings, presented for 15 sec each in fixed order (approx. 5.5 min total).

Social Variable Measures. At pre-treatment, all social variables will be live coded using Pocket Observer installed on tablets. Data will be uploaded to Noldus Observer XT projects on a lab computer for analysis and secondary coding. Research staff will be trained to a criterion of 80% inter-rater reliability on all social behaviors (see Appendix B) over three sessions using recordings of child social interactions from the PIs R01.

Social attention: For the semi-structured (12 min total) and naturalistic assessments (30 min total), social attention will be (a) total % of time child visually attends to the peer's face, and (b) total % of time child visually attends to the adult's face. For eye tracking assessments, social attention will be total % time child visually attends to adult and peer faces in the Static and Dynamic Visual Exploration Tasks.

Peer interest: For semi-structured and naturalistic assessments, peer interest will be total rates of (a) child joint attention to peer actions/activities, by looking at the object being acted on then to peer's face within 5 sec, and (b) child imitation of peer actions/activities within 5 sec. For eye tracking assessments, peer interest will be total % time child visually attends to peer faces and activities during the Interactive Visual Exploration Task.

Baseline and Intervention Outcome Measures. The two primary measures to be collected during baseline and at 8 weeks post start of treatment are: (1) total rate of spontaneous child intentional communication acts, and (2) total rate of reciprocal child-peer exchanges. The time frame was selected based on clear gains observed over the first 8 weeks of treatment in the previous R01 (see Figure 1). For each measure, total frequency will be obtained across three 10-min sessions at each time point (total 30 min pre- and post-), with total acts then averaged across the three sessions and converted to a rate per minute. Change in each rate will be used in the analyses. Intentional acts are defined as the use of any communicative mode (e.g., gestures, speech, SGD) that are clearly directed to a peer using eye contact/body orientation to communicate for a functional purpose (e.g., request objects/actions, comment, or protest). Each intentional act will be coded as an initiation (IN) or response (RS), with RSs coded within 5 sec of the other child's act. If 5 sec pass without a RS, the next act will be coded as a new IN. Each IN + RS sequence between child and peer is coded as one reciprocal exchange; longer IN + RS exchanges are possible if a child continues to respond within 5 sec (i.e., IN + RS + RS + RS - 5 sec pause - IN + RS). This coding system is based on the PIs published studies.

Generalization Outcome Measure. At baseline and 8 weeks post start of treatment, when intervention outcome data are collected, rates of the two dependent variables (i.e., spontaneous intentional child communication acts and reciprocal acts) will be coded for 10 min in two non-intervention settings (total 20 min): snack and free-play. Total rate for each variable will be averaged across the 20 minutes and change in each rate from pre- to post-generalization probes will be used in the analyses.

Baseline Procedures. Prior to treatment, one child-peer dyad will be seated at a table and engage in three 10-min social activities commonly observed in preschool (e.g., matching games, puzzles). Children will be instructed to stay together and play, with the SGD (iPad) programmed to match the activity. No adult prompts will be provided. The two peers recruited for each child with ASD will participate on a rotating schedule.

Intervention Procedures. Four cohorts of children will receive the intervention in two 30-min weekly sessions for 8 weeks, with one cohort starting in the fall and one in the winter of each study year (n=18 Yr 1; n=22 Yr 2). To feasibly intervene with four cohorts, the PI will train research staff to provide the intervention. Using training materials developed in her R01, staff will be trained to 90% fidelity of 10 steps. We will track concurrent speech language therapy, ABA therapy, and other therapies received within and outside of school.

SGD Setup and Familiarity. All children will receive an iPad with Touch Chat app as a SGD to use for the duration of the study. We will recommend all children have SGD access across their school day. School and home SGD usage will be monitored with automated data logging software installed on each iPad.

Peer Training. Peers will be trained in three 30-min sessions following the Stay-Play-Talk with iPad approach from the PIs earlier AAC studies. Each step is taught using a standard protocol of discussion, modeling, practice, and feedback focusing on: (1) Stay (Sit Close; If Buddy Moves you Move), (2) Play (Share Toys; Take Turns), and (3) Talk with iPad (Look and Listen; Push and Talk), and one added step found to increase child responses (4) Get Attention and Hold and Wait. The peer creates an expectant pause for the focal child to communicate by gaining attention, holding an object (out of reach) and waiting.

Direct SGD Instruction with Trained Peer. Each peer will engage in a 15-min activity on a rotating basis. The first 5 min focuses on Stay-Play-Talk skill review, child-peer practice using the iPad and adult feedback. In the next 10 min, the implementer observes the children's interactions and as necessary provides prompts with a least-to-most hierarchy for successful communication exchanges. That is, peers are prompted to model SGD use and help the child push a symbol prior to the adult eliciting child communication. Sessions will be provided twice per week for 8 weeks, a sufficient length of time for intervention response to occur (see Figure 1).

Fidelity Procedures. Two forms will be used to monitor fidelity of adult administration to: (1) coach peers in the pre-treatment semi-structured assessment, and (2) implement twice weekly PMI + SGD treatment. To monitor adherence to the treatment protocol and identify peers who may need more support, peer treatment fidelity will be collected for 30% of sessions. If a peer falls below 60% fidelity over two sessions, a 20-min refresher session will be provided. Mean fidelity of peer implementation in the R01 was 80% for all trained peers, and retraining was necessary with only three peers (5% of total).

Data Analysis. Initial analyses will examine the distributions of the two change in rate variable outcomes, and if necessary, generalized linear models with appropriate link functions will be employed. Initial analyses will examine cohort effects, although no differences are expected. For each aim separate models will be examined for the change in child communication rate and change in rate of reciprocal exchanges. Because of the relatively small sample, limited sets of predictors will be examined in each model. For Aim 1a, separate regression models will be used to determine the relative strength of association and predictive value of each pair of social variables for response to intervention. PLS-5 and DAS-II scores will be entered as covariates into all models because of demonstrated associations between language skills and cognitive abilities and communication outcomes. For example, rate of joint attention and rate of imitation in naturalistic observations will be examined together as indicators of peer interest. Pseudo-R2 values will be compared across the 6 models (peer interest and social attention for each assessment context) for each outcome to identify which social predictors account for the most variance in communication gains. To determine the effect of measurement context on the strength of relationship between social predictors and outcomes (Aim 1b), the two social attention and two peer interest variables from the same context will be examined together but without additional covariates because of the limited sample size. Pseudo-R2 values will be used to descriptively compare the contexts to determine which predictors account for the most variance in intervention outcomes. Multiple imputation or FIML estimation will be used to bring exogenous variables into the likelihood function and account for missing data in both Aim 1a and Aim 1b although missing data was minimal in the PIs previous trial. Parallel analyses will be conducted to determine if social variables predict the two generalized communication outcomes (Aim 2a). Communication outcomes will come from peer interactions in non-treatment settings, but the predictors and approach will be the same as in Aim 1. Similarly, the analyses for Aim 2b will follow the process described for Aim 1b.

Sex as a Biological Variable. Because the sample is likely to be predominantly male due to the gender composition of individuals with ASD, it will not be possible to evaluate sex differences in these effects. The investigators will present results for the total sample, male subgroup, and female subgroup when possible.

Power. With the proposed sample of 40 participants, the investigators will have .80 power to detect moderate associations between social variables and outcomes (f2=.21) in Aims 1a and 1b. For Aim 1b and 2b, we will have .80 power to detect moderate effects of f2=.26.

Inclusion Criteria:

• Children with ASD (N=40) between 3 to 6 years of age; diagnosis confirmed based on DSM-V criteria (APA, 2013) and administration of the Autism Diagnostic Observational Schedule-2 (ADOS-2; Lord et al., 2012).
• minimally verbal defined as less than 20 functional, spontaneous words, signs, or symbols
• currently using or a candidate for a SGD to augment communication
• symbol discrimination skills based on pre-treatment assessment with an adult partner using a matching-to-sample procedure and a minimum choice of one preferred item and one foil
• English as the primary language spoken at home.

Exclusion criteria:

• co-morbid or previously identified genetic conditions associated with ASD (e.g., Fragile X)
• visual impairment, deafness, or other sensory disabilities
• a lack of symbol discrimination skills

Peer Partners:

• ages of 3 and 6 years (N=80 total over 2 years).
• recommended by school staff based on meeting six of the eight following criteria: 1) follows simple 2-step directions, (2) responds positively to adult requests to modify or change undesirable behaviors, (3) appropriately approaches and invites other children to play, (4) is comfortable talking to and interacting with other children while playing, (5) takes turns and shares toys with other children, (6) offers to help other children set up play routines or complete tasks, (7) is well-liked by many children in the classroom, and (8) will stay and play with other children for an appropriate length of time.

The proposed study will include information on child variables that predict response to a peer-mediated intervention, and the impact of different measurement contexts on child performance, for 40 minimally-verbal preschool-age children with autism spectrum disorder (ASD) and 80 peers without disabilities. All data collected will be stripped of identifiers prior to release of sharing data. We will make the data available under a data-sharing agreement that provides for a commitment to use the data only for research purposes and not to identify any participating students or peers. This agreement will also include a commitment to securing the data using appropriate privacy safeguards and computer technology as approved by the Institutional Research Review Board at the University of Kansas, and destroying the data after analyses are complete.

To disseminate the findings of this study, the study results will be reported in clinicaltrials.gov within 1 year of study completion.