Regarding treatment, evidence-based practices recommended for autistic children and adolescents include social skills training (17), frequently implemented in group format in school environments (18). Various social skills programs have been proposed, such as PEERS (19), Think Social! (20), Social Skills Training (21), Navigating the Social World (22) and Social Adjustment Enhancement Intervention (2).

Specific social skills training programs have effectively increased social competence, peer interaction, and functional behaviors while decreasing time alone and repetitive behaviors (23, 24). Research has also found a decrease in overall anxiety (25) and social anxiety (12) after participation in social skills intervention programs. However, some authors (e.g., 26) have suggested that social skills training may also have the potential to exacerbate anxiety, depending on the individual’s mental health profile. Similarly, several studies (e.g., 27, 28) have raised concerns about the potential negative impacts of social skills interventions on autistic individuals. These studies highlight that while such interventions might help individuals conform to socially appropriate behaviors, they can also lead to increased anxiety and emotional distress. Therefore, researchers such as Antshel et al. (29) propose to assess the influence of anxiety on social skill treatment outcomes for children with autism and to investigate whether anxiety would moderate treatment efficacy.

Building on these considerations, certain factors, such as intelligence quotient and age (30), may influence treatment response, but these relationships are unclear. The literature highlights the effectiveness of social skills intervention programs for participants with an IQ above 70 (31), noting explicit improvement in comorbid depression symptoms (2). A systematic review concluded that better cognitive functioning and verbal ability are related to a greater treatment response overall (23). Regarding age, studies suggest more positive results for adolescents than children, with particularly favorable outcomes reported among adolescent females (32). However, there is still considerable controversy about the role of participants’ age in treatment outcomes.

In addition to individual characteristics, contextual and structural aspects of the intervention, such as setting and duration, may also impact treatment outcomes. There have been relatively few studies examining low-intensity social skills programs consisting of limited sessions (typically 10–15 sessions). When implemented in school settings, several studies have demonstrated improvements in participants’ functioning within the school environment (25, 33), enhanced ability to perceive anger (34), reductions in social anxiety symptomatology (12), and an overall improvement in social skills among children and adolescents (35). However, there remains a significant need to evaluate the efficacy of adapted manualized interventions delivered in lower-intensity formats, particularly those designed to meet the needs and constraints of publicly funded community services.

In terms of assessment, greater consensus is needed regarding the most sensitive method to evaluate changes related to treatment response. Standardized questionnaires may not reflect changes in social skills observed during sessions, but they are helpful to assess changes in comorbidity (23, 36). In response to these limitations, researchers have proposed integrating data obtained from questionnaires (quantitative measures) with observational methodologies (qualitative measures) to analyze spontaneous behaviors that occur in response to various social situations and prompts (37). Although further research is needed, observational methodology is increasingly utilized in autism research (38, 39), providing new methods that may contribute to more sensitively capturing the subtle behaviors occurring in social skills development. Along this line, Dean and Chang (40) conducted a systematic review of school-based social skills interventions, specifically examining studies that used observational measures. According to their review, all studies reported improvements in social behaviors, while observation protocols detected more subtle changes.

Recent work by Moody et al. (41) supports the integration of multimodal assessment tools, including behavioral coding systems and ecological momentary assessment, to improve the accuracy of treatment response evaluation in autistic populations. These approaches offer promising directions for future research and clinical practice. The authors also emphasize the importance of moving beyond traditional informant-report measures to include observational and culturally sensitive methods that better reflect the diversity of autistic experiences.

This study was conducted as a randomized, waitlist-controlled pilot efficacy trial to evaluate an adapted version of the Social Adjustment Enhancement Intervention (SAEI) (2), designed for transferability to publicly funded services.

The first step of this study was to develop an adaptation of the social skills training program proposed by Solomon et al. (2): Social Adjustment Enhancement Intervention of the University of California, Davis MIND Institute, a program for autistic children and adolescents as well as those with related communication challenges. The program was designed as a 20-session group program where therapists promote natural and spontaneous social interactions, including reciprocal conversation, cooperative play, and a sense of humor. Previous research found positive results, such as increased socially oriented responding vocalizations and decreased non-social vocalizations (e.g., talking alone) after the intervention (9).

For this study, the research team developed a shorter version of the program to make it feasible in a publicly funded university hospital mental health setting, to facilitate future implementation in community-based services.

Once the program was adapted, the primary goal was to conduct a randomized waitlist-controlled trial to evaluate the effect of our adapted social skills program on autistic children and adolescents. Both observational methodologies and standardized questionnaires were used to assess treatment response, allowing for the examination of real-time social behaviors as well as broader symptom profiles. The study sought to capture the nuanced and immediate changes in social interactions resulting from the intervention, providing a comprehensive understanding of its impact on participants’ social competencies.

Based on this goal, the following research questions were proposed: (1) Do participants in the intervention group demonstrate increased social behaviors (e.g., eye contact, smiling, functional communication, and verbal social interaction) following the intervention, compared to the control group? It was hypothesized that participants in the intervention group would exhibit a higher number of social behaviors in a group situation after receiving the program than the control group (2, 9, 62) (2). What are the differential effects of social skills interventions on anxiety and affective symptoms in autistic individuals, considering both improvements and potential exacerbations? Based on previous research on this program (2), it was hypothesized that participants in the intervention group would show no negative impact on their anxiety or affective problems and might even show improvement in comorbid symptoms after the intervention (12, 25). An exploratory question was also examined: (3) Do treatment outcomes vary by verbal intelligence quotient (IQ) and age group? This question was intended to explore whether the intervention produces differential effects for children versus adolescents, and whether individuals with a verbal IQ above 90 show a more favorable response compared to the full sample.

Psychologists and psychiatrists recruited participants from the Autism unit of a public hospital after being evaluated by highly trained, specialized clinicians in autism. Inclusion criteria were: age range between 8 and 17 years old, a confirmed clinical diagnosis of autism according to DSM 5 (1), meeting the cut-off of Autism at the Autism Diagnostic Observation Schedule-2 (ADOS-2, 42), and a standard range of Verbal Comprehension (Verbal IQ with a cutoff of 80, measured by the Wechsler Intelligence Scale for Children - IV, 2007, 2014). Participants with severe behavioral problems, several mental conditions such as schizophrenia, or intellectual functioning below an IQ score of 80 were excluded. Participants were excluded if they presented intellectual functioning below an IQ score of 80 or a diagnosis of several mental disorders (e.g., schizophrenia, bipolar disorder, conduct disorder or oppositional behavior disorder). All mental health diagnoses were established according to DSM-5 criteria via clinical interview administered by licensed clinical psychologists.

A total sample of 94 participants was divided into two age groups: children (8 to 12 years old, n=36) and adolescents (13 to 17 years old, n=43). Subjects were paired according to age, sex, and verbal IQ; then each pair was randomized between the two groups (experimental group and waiting-list control group) using R software (R Foundation for Statistical Computing; Vienna, Austria) with a fixed random seed (see the Consort Diagram in Figure 1).

Although the full randomized controlled trial (RCT) sample included 94 autistic children and adolescents (i.e., 36 children and 43 adolescents) who completed the social skills intervention, only a subsample of 39 participants met the criteria for inclusion in the observational methodology analysis. These criteria included: 1) attending the specific recorded sessions (i.e., sessions 2 and 10); 2) availability of high-quality video recordings of the sessions, which were required for reliable coding; availability of video recordings with sufficient duration and appropriate camera angles to reliably code the target behaviors, based on audiovisual characteristics (see Figure 1), (3) the presence of observable and analyzable social interaction moments within the group setting. The observational methodology was applied to this subsample to ensure consistent and valid measurement of spontaneous social behaviors. Participants in the full sample were divided into two developmental groups (children and adolescents) and an exploratory analysis was conducted to examine whether those with a verbal IQ above 90 showed a more favorable treatment response compared to the overall sample.

A randomized, waitlist-controlled pilot efficacy trial was conducted between 2017 and 2019 to evaluate the impact of the adapted intervention on social behaviors, internalizing symptoms, and anxiety. Participants were randomized by age, gender, and verbal IQ. The intervention condition was compared at two time points: baseline (T1) and three months later (T2). During these three months between assessments, the experimental group received our adapted social skills intervention program, which consisted of 10 sessions delivered once per week, while the waiting-list control group received the intervention after the T2 assessment.

The social skills intervention program was a consented adaptation of the Social Adjustment Enhancement Intervention of the UC Davis MIND Institute (2). While the original program included 20 sessions and a parent component, the adapted version was reduced to 10 group sessions to improve feasibility in a publicly funded mental health setting. The parent component was removed. Prior community research (e.g., 35, 43) has highlighted the challenges of implementing intensive, evidence-based interventions in these settings, where resource limitations and long waiting lists often necessitate shorter programs that can serve more children. However, However, these adaptations must be rigorously tested before broader community implementation. Importantly, the core components of the original program were preserved.

The original program comprised two modules of 10 sessions: The first module includes topics related to emotion recognition, stress management, and conversational skills, while the second module is centered on social problem-solving and developing a video. The adapted 10-session version retained six sessions on emotional competence (e.g., empathy, recognition, and regulation), two sessions on non-verbal communication and conversation, and one on problem-solving. Problem-solving was also addressed indirectly through real-time support during free play, when challenges naturally emerged.

Each session lasted 90 minutes and was delivered in a group format designed to foster internal motivation for social learning, with individualized goals, cognitive-behavioral strategies, role-play, peer modeling, and clinician-supported opportunities for building friendships. Sessions followed a consistent structure: greeting, free play (unstructured peer interaction), a didactic activity (focused on social competence), joke-telling, and closing activities, which included optional homework (“social experiment for the week”). Didactic content addressed key areas of social competence such as empathy, expressing feelings, and resolving social conflicts.

In addition to following the structure proposed by Solomon et al. (2), clinicians incorporated strategies based on Theory of Mind and social cognition (44). Each session aimed to support the development of social skills frequently impacted in autistic youth (see Table 1 for general structure and Table 2 for session-specific content).

All procedures followed the Declaration of Helsinki 2000 and our hospital’s Ethical Committee for Clinical Research standards. This study was registered in the ClinicalTrials.gov database (Identifier: NCT05713162). Parents of eligible participants were invited to attend an informative session in which the study was presented prior to recruitment. Afterwards, parents provided signed consent, and informed assent was obtained from the child or adolescent.

The study was conducted over three years (i.e., 2017, 2018, and 2019). Each year, the team ran an experimental group of children and another group of adolescents. Treatment sessions were delivered from January to April, and each treatment group consisted of 7 to 9 participants who received the same intervention conditions.

The intervention was delivered to a group therapy room within a pediatric public hospital (review further details in the Intervention Description section). After approval by Solomon and colleagues, the original program materials were translated, adapted, and culturally tailored for local use. A senior psychologist from our department—previously involved in implementing the original intervention at the University of California, Davis, MIND Institute (US), provided training to ensure consistency with the original model. To ensure fidelity to the original program, the clinical team reviewed the implementation protocol and completed a session-specific checklist at the end of each session.

To assess treatment response in social behaviors, researchers’ video-recorded the 14-minute free play period from the second (T1) and last treatment session (T2). The first session was excluded to avoid potential bias in social behavior due to participants meeting for the first time. Free play occurred in an unstructured format, where participants could interact using various games provided or choose to play alone. Therapists and co-therapists interacted only as needed to prompt or observe spontaneous social behaviors, using this segment to track progress. Two cameras were positioned at different angles to record sessions discreetly, following all ethical guidelines. Participants were informed about the purpose, location, and use of the video recordings for research. Two trained psychologists coded all targeted social behaviors. They went through a reliability process in order to assure the quality of the data. Raters discussed codes when there was disagreement. Video coding was performed using LINCE software, a validated multiplatform tool (50). To assess changes in comorbid anxiety and affective symptoms, parents completed standardized questionnaires (CBCL and SCAS) pre- and post-intervention.

Data were analyzed with the Statistical Package for the Social Sciences (SPSS v23.0) (60). Participants were characterized through descriptive analyses. To evaluate whether there was a treatment effect on the number of social behaviors, a 2 X 2 (time x treatment group) ANOVA was conducted for the target observational scale outcomes: eye contact, smile, functional communication, and verbal social communication.

The second objective aimed to evaluate whether participants in the intervention group (experimental group) showed lower levels of internalizing symptomatology at the post-treatment assessment than participants in the control group (waitlist group). This analysis also helped researchers ensure no unwanted side effects of the intervention. To test this, a 2 X 2 (time x treatment group) ANOVA was conducted for specific subscales of the CBCL (anxiety problems, affective problems, internalizing problems) and the parent-report SCAS (social phobia and generalized anxiety) questionnaires.

The last research question asked whether there were differences in the treatment effect for all outcomes based on the group age (children and adolescents) and verbal IQ (VIQ). First, 2 X 2 X 2 (time x treatment group x age group) ANOVA models were performed. Afterwards, researchers selected the sample of children and evaluated significant changes in the experimental group versus the waitlist control group (time x treatment group). The same analyses were replicated for adolescents.

Finally, the study attempted to evaluate treatment response associated with participants’ verbal IQ. To test this question, participants were selected based on their VIQ, using a score of 90 as the cut-off and selection criteria. This allowed detect differences in the treatment responses of participants with a higher VIQ compared to the results obtained with the total sample. For this purpose, researchers conducted a 2 X 2 ANOVA (time x treatment group) within the sample with an IQ over 90 (n=60).

All statistical tests were performed with a bilateral contrast, and the significance level was set at 0.05.

Although a multiple ANOVAs was conducted to test hypotheses concerning social behaviors and internalizing symptoms, no formal adjustment for multiple comparisons was applied.

This decision was made because of the pilot nature of this study and because the analyses were hypothesis driven.

All participants in the total sample (N = 79; 36 children and 43 adolescents) received the entire treatment program. They were 82.3% male (mean age: 11.92; mean verbal IQ: 105.73). Participants in the sub-sample analyzed for observational data (n=39) were 84.16% male (mean age: 11.27; mean verbal IQ: 104.08). Normality and statistical tests were performed to assess whether there were differences between the participants who only had questionnaires as a measure and those who had questionnaires and videos. There were no differences between groups. Descriptive results are shown in Table 4.

Participants, parents, and clinicians reported a high level of verbal acceptability, but no specific rating scale was administered. Clinicians participating in the study showed great satisfaction, as the training did not require expensive materials or specific equipment. Free playtime was highly motivating for participants, especially when they found common interests.

The first research question was whether there were changes in the number of social behaviors in the intervention group between baseline and follow-up. Table 5 shows the means and standard deviations at baseline and follow-up for the experimental and control groups and the ANOVA statistics. Results from the ANOVA did not show significant differences between groups after the intervention; however, the experimental group showed a marginally significant increase in eye contact (F = 3.17; p = .083; effect size = 0.079) and the control group demonstrated a decrease in functional communication (M_Pre= 16.08, M_Post= 11.31). Although these differences did not reach conventional significance thresholds, they may suggest trends worth exploring in future studies. No differences suggesting potential effects were observed for smiling or social verbal communication (see Table 5).

The second research question was whether there were significant changes in comorbid internalizing symptomatology after the intervention, considering both the intervention’s beneficial and potentially harmful effects. Table 6 shows the means and standard deviations of baseline and follow-up for the experimental and control groups and the ANOVA statistics. Results from the ANOVA showed a small, non-significant decrease in the subscale affective problems from the CBCL (F = 2.99; p = .09; effect size = 0.041) in the experimental group. This subscale includes problems such as fatigue, apathy, sadness, and inferiority complex, among others. Other target scales from the CBCL and the Spence Children’s Anxiety Scale did not show any significant or marginal differences (see Table 6), suggesting no significant negative side effects of the intervention.

Based on the previous literature, treatment responses based on age and verbal IQ were examined, replicating analyses on the children group, the adolescent group and the higher IQ subgroup. The subgroup of 23 children (experimental group=14, control group=9) showed significant differences between groups after the intervention in the frequency of eye contact (F = 4.64, p = .043; effect size = 0.181) and functional communication (F = 5.09, p = .035; effect size = 0.195). The experimental group showed an increase in the frequency of eye contact (M_Pre= 0.29, M_Post= 1.86), while the control group diminished the frequency of eye contact (M_Pre= 1.89, M_Post= 0.78). Regarding functional communication, the experimental group showed a slight increase after the intervention (M_Pre= 13.64, M_Post= 14.93), whereas the control group demonstrated a decrease (M_Pre= 19, M_Post= 12.33). Moreover, a difference that approached but did not reach statistical significance between groups was observed for smiles (F = 3.73, p = .067). The experimental group showed an increase in the frequency of smiles (M_Pre= 0.29, M_Post= 1.29), while the control group exhibited a diminished number of smiles (M_Pre= 1.78, M_Post= 0.44). Results did not show differences in social-verbal communication in the group of children (see Table 6).

Results from questionnaires in the group of children revealed significant differences between groups after the intervention in the CBCL internalizing subscale (F = 6.09, p = .02; effect size = 0.16). The experimental group showed a decrease in the score of internalizing problems (M_Pre= 16.17, M_Post= 12.44), whereas the control group maintained a similar mean (M_Pre= 17.75, M_Post= 18.5). Additionally, a small, non-significant reduction was observed for the CBCL affective problems subscale (F = 3.08, p = .09) for the experimental group. No other CBCL or SCAS subscales showed notable differences between groups in the children subgroup, and none of the effects reached statistical significance.

In the subgroup of adolescents, no significant results were observed in either studied variable of the observational scale (n=16). For questionnaires (n=43), results from the ANOVA analyses showed significant differences between groups after the intervention in the parent report subscale of the SCAS generalized anxiety (F = 6.19, p = .02; effect size = 0.15). The experimental group showed an increase in generalized anxiety (M_Pre= 4.33, M_Post= 5.28).

Of the total sample (N = 79), 60 participants had a verbal IQ>90 (experimental group=35, control group=25). Results showed a significant difference between groups in eye contact (F = 4.52; p = .04; effect size = 0.148) and functional communication (F = 5.56; p = .03; effect size = 0.176). The experimental group showed an increase in the amount of eye contact (Children: M_Pre= 0.18, M_Post= 1.55; Adolescents: M_Pre= 3.3, M_Post= 5.3), while the control group showed a decrease (Children: M_Pre= 2.13, M_Post= 0.87; Adolescents: M_Pre= 7, M_Post= 2).

Regarding functional communication, results indicated an increase for the experimental group (Children: M_Pre= 13, M_Post= 14.18; Adolescents: M_Pre= 15.1, M_Post= 16.2), while the control group showed a decrease (Children: M_Pre= 18.38, M_Post= 13.5; Adolescents: M_Pre= 32, M_Post= 13).

No significant differences between groups after the intervention were observed in the results of the questionnaires in the group of participants with VIQ>90, showing no unwanted side effect of the intervention in this higher-functioning subgroup.

First, a limitation of the study was the small sample size for each group and subgroup, which limits interpretations for a broader population and constraints statistical power in the effort to find significance in the tested hypothesis Additionally, the sample included a predominantly male group, which, although representative of clinical prevalence patterns in ASD, limits the generalizability of the results to girls with autism. Socioeconomic and ethnic background data were not analyzed in this pilot trial, but these factors should be carefully considered in future research to ensure broader applicability and equity in intervention outcomes.

Second, a significant limitation was related to the intensity and the program duration, where the total number of sessions, frequency, and/or total duration in time could make a significant difference in our tested hypothesis. Researchers reduced the number of sessions to half of the original twenty to make it available to a larger number of patients that met the initial inclusion criteria for the clinical program and reduce, as much as possible, the waiting time to begin the program.

Third, there were limitations related to measures. Social behaviors were evaluated with an observational instrument in the context of a social skills group, but the study did not include a caregiver scored measure of social interaction, which could have offered information about generalization of skills into other domains of life. However, parents burden is intense regarding participating in a trial so researchers aim to prevent drop-outs by lessening family questionnaires.

Fourth, this observational coding was conducted on a subgroup of participants whose video recordings met criteria for visibility and duration. Due to the use of static cameras in a large room where participants could move freely, some recordings did not adequately capture the child’s behavior and had to be excluded. Although comparisons showed no significant differences between the video-coded subgroup and the overall sample, the exclusion process could represent a potential source of selection bias.

Other limitations included resource constraints within the public healthcare setting where the study was implemented, that lead to an abbreviated cognitive assessment (WISC verbal IQ). As a result, detailed cognitive indices such as verbal comprehension, working memory, or processing speed could not be systematically reported, which limits the granularity of participant characterization and better understanding of the treatment response to targeted subgroups. Also, the lack of parent components and the lack of an independent rater or external therapist to evaluate the “implementation” fidelity, which could be biased by the team leader’s supervision across sessions or age group. Additionally, a waitlist-controlled design was used, and the study did not include follow-up assessments to evaluate maintenance of treatment effects due to the community setting where research was performed.

Finally, the quantity of data collected from the observational methodology was constrained by technological resources (e.g., recording angles) and other factors that limited the sample and the data collection itself, directly affecting the available sample with reliable observational and coded data, in turn affecting the statistical power for the primary objective. No correction for multiple comparisons was applied, which may increase the risk of Type I error. However, this decision was made to preserve statistical power and reduce the risk of Type II error, which is a relevant concern in pilot studies with small sample sizes. Future studies with larger samples should consider multiplicity adjustments to confirm these findings.