DfE review finds no single support model works for all autistic pupils or those with SEMH needs

The analysis of 73 systematic reviews finds the clearest gains in short-term behavior and social communication, while sustained effects in mainstream classrooms remain uncertain

A Department for Education review examines behavioral and social communication interventions for autistic pupils and children with SEMH needs

A July 2026 evidence review published by the Department for Education has found that no single intervention consistently meets the behavioral and social communication needs of all autistic pupils or children with social, emotional and mental health needs in England’s mainstream schools.

The review identifies positive findings for peer support, physical activity, structured behavioral approaches, play-based programs, social skills interventions and some digital tools. However, the strength of the evidence varies, and most studies measure changes over short periods rather than establishing whether improvements last or transfer to other settings and activities.

Published as part of the Department for Education’s wider work on identifying and supporting children with special educational needs and disabilities, the research does not constitute an endorsement of any named intervention or indicate current or future government policy.

Its scope is substantial. During the 2024/25 academic year, approximately 1.44 million pupils with identified special educational needs were attending mainstream state-funded primary and secondary schools in England. This included 1.17 million pupils receiving SEN support and 271,000 with an Education, Health and Care plan.

Among pupils receiving SEN support, 279,230, or 23.6%, had SEMH recorded as their primary need. The equivalent figure among pupils with an EHC plan was 71,304, or 16%.

Researchers Catherine Antalek, Juhayna Taha, Susanne Peter, Aneeza Pervez, Fatema Yusuf Jangbarwala, Daisy Loyd, Amelia Roberts, Matthew P. Somerville and Anna Melissa Romualdez, all from University College London, conducted the rapid evidence assessment.

They searched for systematic reviews and meta-analyses published in English between 2014 and 2025, covering people aged up to 25 with identified autism or SEMH needs. The interventions had to be applicable to mainstream education and capable of being delivered without specialist qualifications.

Of 4,965 records screened, 909 underwent full-text assessment and 73 were included. The PRISMA study selection diagram categorizes 54 reviews as autism research, six as SEMH and 13 as mixed.

The researchers assessed the reviews using the AMSTAR2 quality framework. Most received high or moderate confidence ratings, but some were rated low because of a critical methodological flaw, typically the absence of a formal risk-of-bias assessment.

Evidence is strongest for immediate, observable changes

Peer-mediated interventions were associated with improved social engagement and communication among many autistic pupils, particularly in early years and primary settings. The most consistent results involved directly observed interactions, while evidence for lasting, generalized or academic effects was more variable.

Activity schedules, Check-In Check-Out systems and Daily Behavior Report Cards produced small to moderate improvements in observable classroom behavior, particularly for pupils with attention deficit hyperactivity disorder or those considered at risk of emotional and behavioral difficulties.

Physical activity was linked to reductions in challenging, stereotyped or repetitive behavior and improvements in self-regulation and social-emotional functioning. Longer, more intensive or sport-based programs tended to produce larger effects, but wide differences in activity type, duration and delivery prevented the researchers from identifying an optimal model.

Naturalistic developmental behavioral interventions, play-based approaches, LEGO therapy, structured social skills programs and video-based interventions also produced positive but inconsistent findings. Outcomes varied according to age, initial skill level, intervention intensity, practitioner training and adherence to the intended delivery model.

Evidence for social stories was mixed. Effects were generally small to moderate and appeared stronger when stories were combined with other strategies or followed established implementation criteria.

Digital interventions show targeted rather than universal benefits

A meta-analysis included in the review examined 28 experimental studies of digital interventions involving 1,360 autistic children and adolescents, 1,222 of whom had autism as their primary diagnosis.

The technology ranged from computer programs and mobile apps to DVDs, virtual reality and mixed reality. Sessions lasted between five and 100 minutes, while complete programs ranged from six to 90 sessions, leaving no consistent model of delivery.

The strongest findings concerned social-emotional outcomes, including emotion recognition and social understanding. Evidence was less consistent for language, cognition, physical development and daily living skills. Daily living was addressed by only one study, preventing firm conclusions.

A separate analysis covered six single-case studies of I-Connect, a free mobile and desktop self-monitoring app. The studies involved 14 pupils aged nine to 15 who were receiving additional educational support.

In all six studies, introducing I-Connect was associated with increased on-task behavior. That finding remains narrow, however. Most participants were male, and there was limited evidence that spending more time on task led to better-quality work or sustained learning.

Specialist delivery and pupil voice remain gaps

Implementation presents a practical limitation for mainstream schools. Although the review focused on approaches considered feasible for educators, several included studies involved behavioral therapists, speech and language therapists, educational psychologists, occupational therapists or specialist yoga, mindfulness and arts practitioners.

The researchers state that schools may require additional resources to reproduce those interventions accurately. It remains unclear whether teachers without equivalent training or supervision would achieve comparable results.

Many studies also relied on measures created specifically for the intervention, which may have inflated estimates of effectiveness. Small samples, unclear allocation procedures and inconsistent measurement of delivery quality created further uncertainty.

Most outcomes were reported by teachers or parents, with limited input from pupils themselves. The review cautions that an observable change in behavior does not necessarily establish improved wellbeing, participation or agency. For autistic pupils, narrowly defined behavioral targets may encourage masking or suppression while underlying stress remains.

The authors call for stronger studies conducted in mainstream classrooms, with educators delivering the interventions and pupils contributing their own perspectives. Future research should also test whether benefits persist after an intervention ends and extend beyond the particular behavior, activity or setting being measured.