Which parts of a consultation actually change how clinicians deliver an autism intervention? A new study offers one of the first experimental answers, and the results are strikingly narrow. In a community mental health system serving low-income families, researchers dismantled a group consultation package into its three most common components—case support, skill rehearsal, and performance feedback—and tested each one in isolation. Of the three, only performance feedback was linked to a measurable improvement in clinician skill, and only in a single domain of caregiver coaching. Meanwhile, the consultation program as a whole was associated with a steady weekly rise in the proportion of eligible children receiving the intervention.
The study, published in the Journal of Autism and Developmental Disorders, focused on Project ImPACT, a manualized caregiver-mediated naturalistic developmental behavioral intervention designed to help parents support social communication in young autistic children. Twenty clinicians from six agencies within the Michigan Medicaid Autism Benefit took part. This system funds applied behavior analysis services for low-income youth aged 0 to 21 with an autism spectrum disorder diagnosis, and administrators there were specifically looking for ways to improve caregiver coaching, a practice considered best practice for supporting skill development in autistic children but one that is infrequently delivered in this system. In one region, nearly half of the youth served had not received a caregiver coaching session over six months, and only 2.7% received it at a frequency consistent with evidence-based models.
To isolate the effects of individual training activities, they used a component analysis within a single-case experimental design across agencies. After a two-week initial training period consisting of a six-hour self-directed online course on Project ImPACT content, clinicians participated in twelve weekly group consultation sessions via Zoom, each lasting 1.5 hours. Crucially, the researchers withheld the synchronous advanced workshop that would normally follow the online course, because that workshop involves similar activities to those being tested—role plays and performance feedback—and would have obscured which component produced any observed change.
Each consultation group received all three conditions in four-week blocks, with the order randomized by a random number generator. During case support sessions, the consultant solicited implementation challenges from the group, let clinicians prioritize two or three to problem-solve, and emailed follow-up notes and resources. During skill rehearsal, clinicians completed assigned role plays of coaching session segments, with the consultant playing the caregiver and—importantly—no oral feedback provided afterward, so that rehearsal could be distinguished from feedback. During performance feedback, the consultant screened a five-minute clip from a clinician’s recorded session and led the group through two rounds of commentary, first positive and then corrective. Trained coders who were naïve to condition and timepoint confirmed that the conditions were delivered as intended: case support at 100% fidelity with an average of 39 minutes of active discussion, skill rehearsal at 93% fidelity with about 35 minutes of role play, and feedback at 97% fidelity with roughly 38 minutes of performance feedback.
Fidelity was scored from videotapes of telehealth sessions. Manual adherence, measured on the Project ImPACT Coaching Fidelity Checklist, reflects how closely clinicians follow the manual’s lesson plans, with more central coaching items weighted more heavily; a score of 80% or higher is considered at fidelity clinically. Coaching competency was assessed with the Parent Empowerment and Coaching in Early Intervention measure, which rates four domains—Collaboration, Demonstration, In-Vivo Feedback, and Reflection and Problem-Solving—on a five-point scale, where a score of 4 or 5 per subscale is considered at fidelity. Coders met strict training criteria before coding, scored each video independently, and resolved disagreements by consensus. The analyses retained 154 sessions from 12 clinicians across the three agencies with sufficient baseline data, using two-level multilevel models in which fidelity scores were nested within clinicians.
The headline result was a study in contrasts. Average baseline manual adherence was 62.67%, well below the 80% benchmark, and none of the consultation components moved it: scores in the feedback phase were 4.63 points higher than baseline, a change that was not statistically significant, while skill rehearsal and case support phases were associated with small, non-significant decreases. Competency told a similar story for most domains. Baseline Collaboration averaged 2.57, Demonstration 3.86, and In-Vivo Feedback 2.32—all below or near the fidelity cutoff—and none of the three components produced statistically significant changes in any of these. Skill rehearsal and case support likewise showed no significant effects on any competency domain.
The one exception was Reflection and Problem-Solving, the domain covering reflective listening and open-ended questions that help caregivers evaluate their own use of intervention strategies and troubleshoot obstacles. In the performance feedback phase, estimated scores were 0.32 points higher than in the baseline phase, a statistically significant increase, t(62.12) = 2.65, p = .01. No other component approached significance for this domain. The authors had hypothesized that video-based performance feedback would outperform skill rehearsal, and the results partially supported that prediction—but the improvement was confined to this single competency, and the researchers themselves note it could reflect the consultant’s feedback focusing on reflective coaching skills, or even be a spurious finding in this small sample.
The second aim examined case penetration, defined as the proportion of eligible clients on a clinician’s caseload who actually receive the intervention. Using weekly self-reported questionnaires from all 21 clinicians across all six agencies—288 observations in total—the researchers found that the estimated mean penetration at intake was 24.37%, meaning providers reported using Project ImPACT with about a quarter of eligible clients. During the baseline phase, penetration rose by an average of 0.53% per week, an increase that was not statistically significant. During the twelve-week consultation phase, however, the slope steepened dramatically: penetration climbed by approximately 2.56% per week, a statistically significant increase, t(275.29) = 3.45, p < .001. Notably, clinicians were never explicitly instructed to offer the intervention to additional clients, which the authors suggest indicates that consultation support itself encouraged clinicians to extend the intervention to more eligible families—possibly through increased self-efficacy, the heightened salience of frequent consultation, or social pressure from observing peers also implementing the intervention.
The authors are careful about the limits of these findings. This was a small pilot study with limited statistical power, and they explicitly state that the lack of significance in the null results does not necessarily indicate a lack of evidence—fidelity changes might emerge in a larger sample. Only seven follow-up sessions were submitted after consultation ended, too few to analyze. The fidelity measures lack published psychometrics, and their cutoffs have not been empirically derived. Nearly all models produced a singular fit, meaning random effects were too small and complex for the dataset, though the authors note this rarely biases the fixed-effect estimates that addressed the primary question. By chance, all three agencies retained for the fidelity analyses began consultation with skill rehearsal, and two shared the identical component order, limiting variation for that condition. The study also did not correct for multiple analyses, and no autistic people were involved in the design or implementation, as the authors acknowledge.
Still, the study breaks methodological ground. To the authors’ knowledge, it is the first empirical study to examine active ingredients of consultation by experimentally manipulating sessions so that each contained only one training activity at a time, and the first to investigate how consultation dosage—weeks of sessions—affects case penetration over time while holding pre-consultation levels constant. The practical takeaway for under-resourced systems is targeted: if the goal is to raise the quality of caregiver coaching, performance feedback using video review deserves emphasis in consultation, while case support and rehearsal without feedback show insufficient evidence as stand-alone active ingredients at this dosage. Because combinations of activities likely outperform any single element, the authors call for future studies using sequential multiple assignment randomized trial designs to identify which sequences of training activities work best, and for replication of the feedback effect at higher doses and in larger samples.
Subject of Research: Experimental testing of which discrete consultation components improve clinician fidelity and case penetration for a caregiver coaching autism intervention in a community mental health system
Article Title: Examining the Active Ingredients of Consultation: Effects of Discrete Consultation Components on Implementation Outcomes for an Autism Intervention in a Community Mental Health System
Article References: Straiton-Webster, D., Moeyaert, M., Greatorex, J., Ewell, A., & Ingersoll, B. (2026). Examining the Active Ingredients of Consultation: Effects of Discrete Consultation Components on Implementation Outcomes for an Autism Intervention in a Community Mental Health System. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-026-07536-5
Image Credits: AI Generated
DOI: 10.1007/s10803-026-07536-5
Keywords: autism intervention, Project ImPACT, consultation, performance feedback, clinician fidelity, case penetration, caregiver coaching, community mental health, Medicaid, implementation science, single-case design, telehealth
News Source: Ophelia Keating. (November 4, 2026). Performance Feedback Emerges as Key Driver of Autism Intervention Training in Community Clinics. Scienmag.
