Classroom-based peer interventions targeting autism ignorance, prejudice and/or discrimination: a systematic PRISMA review

ABSTRACT

A growing movement towards inclusive education worldwide means more children with autism are being educated in mainstream classrooms. However, integration does not equate to inclusion, and peer stigmatisation of autism is commonplace. This systematic review explores the merit of school-based peer education interventions targeting the stigmatisation of children with autism spectrum disorders. Relevant records published from 1994 were identified through systematic searches of five electronic databases: EBSCO, Medline, PsychInfo, PsychArticles and ERIC. Thirty-one documents pertaining to 27 studies met the pre-specified eligibility criteria and were included for data-extraction. A narrative synthesis highlighted significant flaws in the available literature, most notably poor methodological quality and a narrow research focus with regards to age, the gender of target child and implementation methods. Nevertheless, this study reports on evidence tentatively supporting the efficacy of ASD de-stigmatisation interventions when targeting ignorance and prejudice. Although no one approach can be determined as most effective, manualised programmes combining different types of information about ASD and delivered using various mediums seem to hold the most promise.

KEYWORDS:

• Autism
• stigma
• inclusive education
• intervention
• classroom
• ignorance
• prejudice
• discrimination
• systematic review
• PRISMA

Introduction

The publication of the United Nations Educational, Scientific and Cultural Organization’s (UNESCO) Salamanca Statement represents a clear shift away from the acceptance of segregated education. This influential document has made inclusive education central to human rights, and a policy objective in most countries (United Nations Educational Scientific and Cultural Organisation 1994). Since there has been an increasing worldwide social impetus to educate those with disability/disabilities alongside their peers without disabilities in mainstream education settings. Yet, inclusion policy has to date focused predominantly on the promotion and facilitation of integration (i.e. a physical presence in mainstream education settings).

As an example, in response to the Salamanca Statement, the Irish government passed the 1998 Education Act (Government of Ireland 1998) which obliged schools to cater for the educational needs of all children in Ireland. Later, the Education for Persons with Special Needs Act (Government of Ireland 2004) made several specific provisions for the education of Irish children with disability/disabilities including the right to be educated in a mainstream school. However, full inclusion remains to be realised. A withdrawal model of additional support (or integrated segregation), involving the removal of children with disability/disabilities from their regular education classroom for one-to-one/small group work is commonplace in Ireland (Ware et al. 2011). Similarly, UK research suggests that only 32% of children with disability/disabilities fully attend their regular mainstream class (Croll 2001). In addition, and counter to the inclusion movement, the use of segregated classrooms (or ‘units’) attached to mainstream schools is increasing in Ireland. Student enrolment in these settings has almost doubled from 2004 to 2018 (Department of Education and Skills 2019). Therefore, despite a legislative shift towards inclusion, educational practices would appear to be more in line with integration.

It is argued that, in addition to a physical presence, full inclusion encompasses acceptance (by both teachers and peers), participation (quality experiences of the social, emotional and academic aspects of education within the school community), achievement (academic progress and improvements in social and emotional skills), and prevention of ‘integrated segregation’ (e.g. the use of withdrawal classes; Booth and Ainscow 1998). While children with disability/disabilities have increased opportunity for physical integration into mainstream schools, there is no certainty of experience of this broader conceptualisation of inclusion.

This is particularly true of individuals with a diagnosis of autism spectrum disorder (ASD),¹ the characteristics of which can impede social inclusion in general education settings (Humphrey and Symes 2011). Autistic individuals universally experience difficulties with social communication and in social development, and often display ritualistic and stereotypical behaviours, resistance to change, and sensory sensitivities (American Psychiatric Association 2013). These social, communication, and behavioural difficulties place this population at an increased risk for social exclusion compared to children with none and/or other disabilities (Campbell et al. 2017; Hebron, Humphrey, and Oldfield 2015; Humphrey and Symes 2010, 2011; J. Schroeder et al. 2014a).

Research suggests a high prevalence of stigmatisation of children with ASD by their peers in integrated mainstream school settings. Stigma encompasses three elements: (1) ignorance (inaccurate knowledge); (2) prejudice (negative attitudes); and (3) discrimination (negative behaviours or behavioural intent; Thornicroft et al. 2007). Although peer awareness of autism is reportedly growing with increasing numbers of autistic children being educated in mainstream schools, studies suggest this awareness is basic (e.g. self- reporting that they know what autism is, or what the term autism means; Campbell and Barger 2011; Campbell et al. 2019; Dillenburger et al. 2017). Although aware of autism, peers continue to have inaccurate knowledge of autism: being unable to accurately define what autism means (Campbell and Barger 2011; Campbell et al. 2004, 2005; Dillenburger et al. 2017; Magiati, Dockrell, and Logotheti 2002); and are ignorant of autism as a disability associated with social communicative and behavioural difficulties (Scheil, Bowers-Campbell, and Campbell 2017).

Misinterpretation and misattribution, stemming from this ignorance, of behavioural manifestations of autism can result in negative attitudes towards autistic peers and discrimination in the form of social exclusion. Indeed, compared with their typically developing peers, research suggests that autistic students have reduced social interactions, spend less time in cooperative activities and more time in solitary activities, and have significantly fewer high-quality friendships (Bauminger et al. 2008; Humphrey and Symes 2011; Kasari et al. 2011; J. J. Wainscot et al. 2008b). Reflecting this, students with ASD in mainstream classes have been found to self-report fewer friendships, lower levels of social support, high rates of loneliness and low social acceptance (Bauminger et al. 2008; Campbell et al. 2019; Chamberlain, Kasari, and Rotheram-Fuller 2007; Connor 2000; Humphrey and Symes 2011; Osler and Osler 2002; Symes and Humphrey 2010).

Of concern is that low social acceptance increases the risk of victimisation in the neurotypical population (Godleski et al. 2015; Veenstra et al. 2007, 2010); and research suggests that for children with autism there is a four-fold likelihood of experiencing bullying (Campbell et al. 2017; J. Wainscot et al. 2008a), and an increased likelihood relative to other disability groups (e.g. dyslexia; Rowley et al. 2012; Symes and Humphrey 2010). The consequences of bullying for the autistic population are serious, with research suggesting it is associated with negative academic, social, and psychological outcomes (J. H. Schroeder et al. 2014b).

While the use of school-based interventions targeting the social communication difficulties of autism has had undoubted benefits for these children (see Sutton et al. 2019 for a systematic review of existing literature), their effectiveness will always be limited given inevitable residual socialisation difficulties. Despite some social skill improvement within the autistic individual, social interaction requires the acceptance of another. Thus, the degree to which children with autism are stigmatised by their peers within mainstream school settings is cause for concern and has led many to suggest the focus of intervention should be widened beyond the individual with ASD to include their peers and the larger school system (Sutton et al. 2019).

In response, the development and evaluation of school-based interventions aimed at reducing the stigmatisation of children with ASD have gathered momentum. Peer-directed educational approaches generally include the provision of all, or a combination of descriptive information designed to promote likability by highlighting similarities between a student with ASD and the peer group; explanatory information to provide attribution for deviant behaviour; directive information which provides suggested behaviours to better include autistic peers; and ASD facts providing information about the characteristics of ASD. A review of the theory underlying each of these forms of information has been outlined in previously published works (see for example Ranson and Byrne 2014). Thus, these educational interventions are intended to target stigma through the provision of accurate, age-appropriate and relevant information about ASD, as well as explicit strategies to better understand and interact with individuals with ASD (Scheil, Bowers-Campbell, and Campbell 2017).

Research aims

This systematic review of the available literature aims to provide a comprehensive overview of the merit of school-based educational approaches to de-stigmatisation of ASD through the exploration of the following questions:

1. How are experimental evaluations of peer-directed ASD de-stigmatisation interventions applied to school settings? Specifically, what interventions are used, how are they implemented, and what research design is used to evaluate their merit?

2. What is the overall quality of identified studies that have both included and evaluated an ASD de-stigmatisation element to intervention as measured by published quality rating scales?

3. Do the results of identified studies support the use of peer education to target the stigma experienced by schoolchildren with ASD?

Methods

We performed a systematic review in accordance with the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) guidelines. The methods were predetermined and specified in a protocol submitted to the PROSPERO database of systematic reviews in October 2019.

Search strategy

Relevant records were identified through systematic searches of five electronic databases: Academic Search Complete (EBSCO), Medline, and the PsychInfo, PsychArticles and Education Resources Information Center (ERIC) databases through the Proquest platform. With input from a social sciences librarian, the PICOS framework was used to help structure and identify relevant terms for searches. In all databases, the following search term combinations were entered in the subject terms (SU) field: (Autis* OR pervasive OR Asperger*) AND su(stigma* OR accept* OR attribution* OR attitude* OR prejud* OR knowledge OR ignoran* OR discriminat* OR aware* OR initiation* OR contact* OR misperce* OR perception* OR belief* OR bully* OR victim* OR friend* OR intent* OR view* OR judgement*) AND su(intervention* OR programme* OR educat* OR train* OR inform*) AND su(school* OR preschool* OR mainstream* OR primary OR secondary OR kindergarten OR elementary OR crèche OR Montessori OR early year* OR classroom*) AND su(peer* OR child* OR adolescen* OR student* OR pupil*). No ‘language of publication’ or ‘publication type’ limiters were applied. In line with the study aims, when possible, database searches were limited by population (human), age (to include those individuals 18 years or younger) and limited to documents published from 1994 (following the publication of the UNESCO Salamanca Statement).

After removing duplicates, all retrieved documents were screened independently for relevance against the eligibility criteria by ?.?(first author) and ?.? (third author), first by title and abstract and then when indicated by full-text. There was substantial agreement (98%; Cohen’s κ = .687) between screeners for title and abstract, and perfect agreement (100%; Cohen’s κ = 1) for full-text screening. Disagreements were resolved by discussion and shared consensus. To identify additional relevant publications, electronic searches were supplemented by the manual search of reference lists of included studies and all their authors. This was an iterative process until no more eligible documents were sourced. This method of document identification is depicted by the PRISMA flow diagram in Figure 1. Searches of databases, reference-lists and author publications occurred between October 2019 and January 2020 and were conducted by SM. Covidence software was used to manage references and record progress (Covidence systematic review software).

Figure 1. PRISMA flow diagram of document identification.

Eligibility criteria

To be included in this review, studies were required to meet a set of pre-determined inclusion criteria. Requirements for inclusion were: (1) population – must target peer stigmatisation of children with ASD. Given the search period (1994 onwards), acceptable population descriptive terms included autism, autistic disorder, high functioning autism, low functioning autism, Asperger’s syndrome, atypical autism and pervasive developmental disorder not otherwise specified, to reflect the varying diagnostic labels used over time (American Psychiatric Association 1980, 1994, 2013); (2) intervention – must describe the provision of stigma related information to peers (to include one/a combination of descriptive, explanatory, directive, and/or ASD facts); (3) comparison and study – explores, using any comparison (including subjective reports) and using any methodology, the impact of this intervention on peer stigma; (4) outcome – reports on the impact of intervention on stigma variables, including knowledge, attitudes and/or social behaviours (intended or actual); and (5) delivers the intervention through a school/classroom catering for children under 18 years of age (including pre-schools, primary and secondary schools).

Additionally, studies were excluded if: (1) data relating to the impact on peers could not be isolated and extracted (i.e. the outcomes measured change in the ASD population only); (2) data relating to the stigma of individual(s) with ASD specifically could not be isolated and extracted if the intervention targets stigma of disability/difference more generally; (3) the intervention targeted stigma in a different population other than peers (e.g. teachers); and (4) interventions were delivered through third level institutions. When possible, authors of included studies were contacted to obtain information that would support a study’s inclusion e.g. to source autism specific data from a study which published data on a general disability population.

Assessment of methodological quality

Relevant outcomes (for this review) for each of the included studies were rated for methodological quality using published rating scales. Outcomes assessed using qualitative methodologies were judged for rigour using the NICE quality appraisal checklist for qualitative studies (National Institute for Health and Care Excellence 2012b). Each study’s methodology was judged to be either of high, low or unclear quality for each of the domains: theoretical approach, study design, data collection, trustworthiness, data analysis, ethics, and overall assessment. For quantitative outcomes, methodological quality was assessed using the NICE Quality Appraisal Checklist for quantitative intervention studies (National Institute for Health and Care Excellence 2012a). This checklist evaluated each study’s internal and external validity as either high, medium and low risk by appraising: characteristics of the participants, the definition of and allocation to study conditions, outcomes over time, and methods of analyses. This tool provided well-defined instructions on how to rate each criterion and allowed for the systematic evaluation of studies with a range of quantitative experimental designs. Risk of bias was further assessed using the Cochrane recommended tool ‘Risk Of Bias In Non-randomised Studies – of Interventions’ (ROBINS-I; Sterne et al. 2016). Using the checklist guidelines, each study assessing a quantitative outcome was appraised for risk of bias in the following domains: confounding, participant selection, classifications of interventions, deviations from intended interventions, missing data, measurement of outcomes, selection of reported results, and overall bias. This process of double assessment of quantitative outcomes ensured that each study was assessed consistently, which limited reviewer bias in the assessment of quality.

Data extraction

When a study was described by more than one included report, data was extracted from all documents for that study. Records meeting all inclusion criteria were coded to extract data relating to population (e.g. sample size, participant demographics, target ASD child characteristics), key characteristics of intervention (e.g. content, delivery), evaluation design, outcome measures, and results.

Analysis

There was significant methodological and clinical heterogeneity between included studies. A meta-analysis of quantitative data was viewed as inappropriate and a descriptive approach to synthesis was employed. Observed patterns of similarity and difference between study populations, methodologies and results were critically appraised. Studies rated as high methodological quality were analysed to assess the utility of peer targeted ASD de-stigmatisation interventions and to identify factors that are likely to support or hinder effectiveness.

Results

Study identification

Figure 1 depicts a PRISMA flow diagram detailing the results of the literature search. The searches retrieved 1,502 documents, 1,286 following the removal of duplicates, which were screened for title and abstract. A list of excluded documents is available from the first author by request. A total of 47 documents were subject to a full-text screen for eligibility, of which 23 were excluded (reasons for exclusion at this phase are included in Figure 1). Supplementary searches of the reference lists and authors of all included studies resulted in the identification of an additional seven studies. A total of 31 identified documents met eligibility criteria describing 27 studies: Campbell et al. (2005) and Campbell et al. (2004) were articles describing the same intervention with the same participants; while the pairings of Ezzamel (2016) and Ezzamel and Bond (2017), Sreckovic (2015) and Sreckovic, Hume, and Able (2017), and Swaim (1998) and Swaim and Morgan (2001) were all combinations of a published thesis and a peer reviewed article.

How are experimental evaluations applied in terms of approach to intervention and research design?

Table 1 reports on the characteristics of the 27 studies included in the final review, and Table 2 summarises these key characteristics. Most documents were peer-reviewed journal articles (80.7%), with the remainder published theses. Studies were conducted in the USA (55.6%), the UK (25.9%), Australia (7.4%), Greece (7.4%) and in Israel (3.7%). All studies focused on stigma in ASD specifically, with the exception of Frederickson, Warren, and Turner (2005) which focused on disabilities more generally but ASD data could be isolated and extracted.

Table 1. Characteristics of studies assessing school-based peer targeted ASD stigma interventions.

Authors (date) Setting	Sample characteristics	Participant Eligibility criteria	Target ASD child(ren) characteristics	Outcome(s) of interest	Method of assessment Data collector	Evaluation design
Campbell (2007) Public middle school USA	n = 233 140 females; mean age 13.07 (SD = 1.01); 20 homeroom classes; African American (45.1%), European American (32.6%), Hispanic/Latino (5.6%), Asian American (1.3%), other (14.2%).	None reported	n = 1 unfamiliar Hypothetical 14-year-old male peer who engaged in autistic-like behaviour	Knowledge, attitudes, behavioural intentions	Heard of autism yes/no question; Knowledge of Autism Questionnaire (Campbell 2007); Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977); adapted Shared Activities Questionnaire (Morgan et al. 1996); Similarity Rating Form; Perceived Responsibility Questionnaire Researcher	Randomised post-test control group
Campbell et al. (2019) Elementary schools (n = 3) USA	n = 234 Fourth- (n = 153) and fifth-graders (n = 81); 19 classes; 110 females; mean age 9.86 years; White (n = 122), Black/African American (n = 31), Asian American (n = 14), other (n = 36), Hispanic (n = 33)	None reported	n = 1 unfamiliar Hypothetical 14-year-old male peer who engaged in autistic-like behaviour	Knowledge, attitudes	Adapted Knowledge of Autism Questionnaire – Modified (Campbell and Barger 2011); Children’s Attitudes Toward Children with Handicaps – Modified (Rosenbaum et al. 1986) Researchers	Pseudo-randomised post-test with control group and 1-week follow-up
Campbell et al. (2004, 2005) Public elementary schools (n = 5) USA	n = 576 third- (n = 194), fourth- (n = 172) and fifth-graders (n = 210); 31 regular education classes; 282 females; age range 8–12.5 years (M = 10.06, SD = .98); African American (8.9%), Caucasian (80.6%); Hispanic/Latino (4.5%), Asia-American (0.7%) and other (5.2%)	In classes without ASD peers; had not heard of autism/ understood autism pre-intervention	n = 1 unfamiliar Hypothetical 12-year-old male peer who engaged in autistic-like behaviour	Attitudes, behavioural intentions	Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977); 12-item version of the Shared Activities Questionnaire (Morgan et al. 1996); Similarity Rating Form Researchers	Pseudo-randomised post-test between groups
Collet-Klingenburg, Neitzel, and LaBerge (2012) Rural middle school USA	n = unknown, (5–7 per target) Sixth/seventh grade	In >1 class with target; interest in group work supporting students; previously demonstrated positive interactions with target	n = 2 familiar 12-year-old, verbal male, partially integrated (data on attitudes and behaviours); 13-year-old, verbal female, partially integrated (data on behaviours)	Attitudes, behaviours	Focus group; observations of peer initiations Researchers	Mixed method: one-group repeated measures; qualitative focus group
Cook (2017) Rural middle schools (n = 2) USA	n = 8 5 females; 4 sixth- and 4 ninth-graders	Previous involvement with LINKS; in class with target; grade B or better; had no discipline referrals	n = 4 familiar All male; 2 × 12-year-olds, 2 × 16-year-olds; 2 sixth- and 1 ninth-grader, 1 in special education class; 2 Caucasian, 2 Hispanic	Attitudes	Participant interviews Researcher	Thematic qualitative interviews
Ezzamel (2016), Ezzamel and Bond (2017) Mainstream primary school UK	n = 5 Aged 7/8; 2 females	Attend school committed to inclusion; in class with target; regular attendance; good social and communication skills; good class participation; teacher recommended; popular	n = 1 familiar Male aged 8 in Year 3. Fully integrated.	Attitudes, behaviours	Observations of peer initiations and responses; focus group, target child and school staff interviews Researchers	Mixed method: one-group repeated measures with 4-week follow-up; thematic qualitative
Fleva (2014) State mainstream (secondary) school UK	n = 179 82 females; age range 10–16 (M = 13.7)	In class without a child with ASD	n = 1 unfamiliar Hypothetical male described as similar to participants in age	Attitudes, behavioural intentions	Adapted Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977); adapted Shared Activities Questionnaire (Morgan et al. 1996) Not reported	Pseudo-randomised post-test between groups
Fleva (2015) Mixed higher secondary (n = 1) and mixed lower secondary (n = 1) schools Greece	n = 416 Mixed gender; attended either Year 7 (n = 91, mean age 12.3) or Year 10 (n = 88, mean age 15.3)	In class without a child with ASD	n = 1 unfamiliar Hypothetical male described as similar to participants in age	Attitudes, behavioural intentions	Translated Shared Activities Questionnaire (Morgan et al. 1996); Asperger Syndrome Questionnaire (adapted from the Chedoke-McMaster Attitudes Towards Children with Rosenbaum et al. 1986) Teacher	Pseudo-randomised post-test between groups
Frederickson, Warren, and Turner (2005) Mainstream (primary) school UK	n = unknown	None reported	n = 1 familiar Male, aged 7 years 11 months	Attitudes	The Like to Play Questionnaire from the Social Inclusion Survey (Frederickson and Graham 1999); Guess Who peer activity (Frederickson and Graham 1999) Trained assistant educational psychologist	One-group repeated measures 18-week follow-up
Gardner et al. (2014) High schools (n = 2) USA	n = 6 4 females; 3 twelfth-, 2 eleventh- and 1 tenth-grader; 4 Hispanic, 2 European American	Good social skills; judged likely to get along with target	n = 2 familiar 18-year-old European American male in senior year, partially integrated; 14-year-old Hispanic male with an intellectual disability, in a ninth-grade special education class.	Attitudes	Bespoke Likert scale questionnaire Self-report	One-group post-test
Gus (2000) Secondary school UK	n = unknown Year 10 classmates	None reported	n = 1 familiar Male in mainstream Year 10 class, receives 1:1 paraprofessional support	Knowledge, attitudes, behaviours	Bespoke feedback questionnaires with open-ended questions; bespoke follow-up questionnaire Self-report	Post-test open-ended qualitative questionnaire 23-week follow-up
Houston (1998) Elementary school USA	n = 6 Education intervention – 2 males, 2 females; Mean age 7 years 8 months; 3 Caucasian, 1 Black; 2 each from first- and third-grades. Social training only: 2 Caucasian males, mean age 5 years 9 months, in Kindergarten	Regular attendance; appropriate and frequent peer interactions; compliant; ability to follow through on lengthy tasks	n = 3 familiar All IQ>70, Caucasian and fully included. Students involved in intervention: 7-year old female in first-grade, 9-year-old male in third-grade. Control: 7-year-old male in Kindergarten	Behaviours	Observations of peer initiations and responses School psychologist and special education teacher	AB
Hughes et al. (2013) Urban public high school USA	n = 3 Male general education students; 2 Black, 1 Caucasian; 2 seniors, 1 sophomore	Shared class with target; judged likely to interact with students with additional need; expressed willingness to interact with target; agreement to self-monitor and record interactions	n = 3 familiar 16-year-old male, Black, sophomore partially integrated; 17-year-old female, Caucasian, junior with severe intellectual disability partially integrated; 17-year-old male, Caucasian junior with mild intellectual disability partially integrated.	Attitudes, behaviours	Observations of peer initiations; qualitative interview Graduate students in special education and researcher	Mixed method: Multiple baseline across participants; qualitative interview
James (2011) Mainstream (primary) school UK	n = 6	None reported	n = 1 familiar Male aged 10	Attitudes	Participant interviews coded using an adaptation of the Coding Scheme of Perceived Causality (Elig and Frieze 1979) Researcher	One-group repeated measures
Mavropoulou and Sideridis (2014) Urban elementary schools (n = 10) Greece	n = 475 200 females; Fourth-, fifth- and sixth-graders; mean age 10.97 (SD = .85); 22 classes	None reported	n = 15 familiar n = 1 unfamiliar One hypothetical peer (no information) Familiar peers: 13 males; mean age 10.71 (SD = 16.85); verbal or alternative communication skills; from public special schools	Knowledge, attitudes, behavioural intentions	Adapted Knowledge of Autism Questionnaire (Ross and Cuskelly 2006); adapted Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977); Shared Activities Questionnaire (Morgan et al. 1996) Researcher	Post-test between groups 3-month follow-up
Morton and Campbell (2008) Public elementary schools (n=5) USA	n = 296 141 females; 20 regular classes; age range 8.42–12.42 (M = 10.21, SD = 0.88); third-, fourth- and fifth-grader; 9.8% African American, 82.8% Caucasian, 4.4% Hispanic/Latino, 0.3% Asian-American	Had not heard of/understood ASD pre-intervention	n = 1 unfamiliar Hypothetical 12-year-old male peer who engaged in autistic-like behaviour	Attitudes, behavioural intentions	Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977); 12-item short version of the Shared Activities Questionnaire (Morgan et al. 1996) Researcher	Pseudo-randomised post-test between groups
O’Connor (2016) (Primary) school UK	n = unknown Classmates representing a wide range of abilities	None reported	n = 1 familiar 10-year-old male	Attitudes	Social Inclusion Survey (Frederickson and Graham 1999) Not reported	One-group repeated measures
Owen-DeSchryver et al. (2008) Public integrated (elementary) schools (n = ?) USA	n = 32–62 Intervention condition – 5 boys, 7 girls; 6 s-, 3 fourth-grader; Control condition – no demographics	Regular attendance; compliant; ability to make up missed schoolwork	n = 3 familiar 7-year-old male fully integrated with a FSIQ of 76; 10-year-old male partially integrated with a FSIQ of 97; 7-year-old male fully integrated and received paraprofessional support within the classroom.	Behaviours	Observations of peer initiations Graduate students of clinical psychology	Multiple baseline across participants, with a between groups element
Ranson and Byrne (2014) Mainstream independent all-girls Catholic school Australia	n = 273 All female; 116 seventh-, 104 eighth-, and 53 ninth-graders; n = 48 intervention condition were two classes in the eighth grade (median age 13 years, range 12–13 years). Controls were all other classes (median age 12 years, range 12–15)	None reported	n =  unknown measures answered with reference to ‘their peers with autism’	Knowledge, attitudes, behavioural intentions	Autism Knowledge Questionnaire (Staniland and Byrne 2013); Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977); adapted Shared Activities Questionnaire (Morgan et al. 1996); Similarity Rating Form Revised (Staniland and Byrne 2013); Perceived Responsibility Questionnaire Revised (Staniland and Byrne 2013); Student Interaction Questionnaire Revised (Staniland and Byrne 2013); student feedback survey Researcher	Pseudo-randomised repeated measures control group 3-month follow-up
Reiter and Vitani (2007) Mainstream primary school Israel	n = 23 11 females; fourth grade; aged 9/10 (average age 9)	None reported	n = 2 familiar High functioning and verbal	Attitudes	Children with Handicaps-Modified (Rosenbaum et al. 1986) Not reported	One-group repeated measures
Scheil, Bowers-Campbell, and Campbell (2017) Medium sized (elementary and middle) schools (n = 2) USA	n = 16 10 females; age range 10–15 (M = 12.3, SD = 1.5); 8 fifth- and 8 eight-graders from 5 classes	In class without a child with ASD	NA	Knowledge	Semi-structured interviews Graduate students	Grounded theory qualitative
Silton and Fogel (2012) Jewish day elementary schools (n = 4) USA	n = 158 26 fourth-, 80 fifth-, and 52 sixth-graders. Mean age 10.39 years (SD = 0.97). 51.9% male. All were Caucasian except for one Asian participant. 76.6% of participants self-reported having heard of autism before intervention.	In class without a child with ASD	n = 3 unfamiliar Before intervention- measures completed with reference to ‘a child with a special condition called autism’; after intervention with reference to two children in video vignettes (one male, one female)	Attitudes, behavioural intentions	Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977); Shared Activities Questionnaire (Morgan et al. 1996) Researcher	Pseudo-randomised repeated measures between groups
Simpson and Bui (2016) Public elementary school USA	n = 12 Second graders from the same class; aged 7/8	None reported	n = 4 familiar 3 male; aged 5–8 years; same class; all FSIQ <80; 3 Caucasian, 1 Asian	Attitudes, behaviours	Semi-structured interviews; observations of peer initiations and responses Researcher and trained paraprofessionals	Mixed method: reversal ABAB single subject; thematic qualitative
Sreckovic (2015), Sreckovic, Hume, and Able (2017) Rural high school USA	n = 14 9 females; aged 14–16 years; ninth- (n = 7), tenth- (n = 6) and eleventh-graders (n = 1); 5 African Americans, 1 Hispanic, and 8 Caucasians; all involved in extracurricular activities	Expressed willingness to develop relationship with target; commitment of one semester; judged socially able; in same lunch period as target; nominated by target as preferred peer and/or had similar interests or group memberships to target	n = 3 familiar 15-year-old multiracial male in the 10th grade, partially integrated; 15-year-old Caucasian male in 9th grade fully integrated; 15-year-old Caucasian male in the 10th grade, partially integrated.	Attitudes, behaviours	Observations of peer initiations; bespoke sociometric measure Researcher and research assistant	Multiple baseline across participants
Staniland and Byrne (2013) Mainstream independent all-boys Catholic school Australia	n = 395 All males; 146 seventh-, 112 eighth-, and 137 ninth-graders; n = 46 intervention condition from 2 eighth-grade classes	Classes with a high number of children with autism could not be in the intervention condition but could be in the control groups	n = unknown measures answered with reference to ‘their peers with autism’	Knowledge, attitudes, behavioural intentions	Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977); adapted Shared Activities Questionnaire (Morgan et al. 1996); bespoke Autism Knowledge Questionnaire; bespoke Similarity Rating Form Revised; bespoke Perceived Responsibility Questionnaire Revised(Staniland and Byrne 2013); bespoke Student Interaction Questionnaire Revised Researcher	Pseudo-randomised repeated measures control group 1 term follow-up
Swaim (1998), Swaim and Morgan (2001) Public (elementary) schools (n = 2) USA	n = 233 117 females; 112 third-graders (mean age 9.12 years) and 121 sixth-graders (mean age 12.02 years); two schools (n = 115 and n = 118); 93.6% Caucasian and 6.4% Black	In class without a child with ASD	n = 1 unfamiliar Hypothetical 12-year-old male peer who engaged in autistic-like behaviour	Attitudes, behavioural intentions	Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977); Shared Activities Questionnaire (twice, once unaltered and once modified to report on perceived classmates’ intentions Morgan et al. 1996); Similarity Rating Form Researcher and research assistants	Pseudo-randomised post-test between groups
Whitaker et al. (1998) Mainstream schools (n = 5) and school for pupils with moderate learning difficulties (n = 1) UK	n = 52 a ‘gender balance and maintaining diversity’ Included some children with emotional and behavioural difficulties	Caring and competent	n = 6 familiar Aspergers or higher functioning autism	Attitudes, behaviours	Interview and bespoke questionnaire Adults not involved in intervention	Mixed method: one-group post-test; qualitative interview

Table 2. Number of studies (and percentages) reporting the listed characteristics.

		n	%
Country
	USA	15	55.6
	UK	7	25.9
	Australia	2	7.4
	Greece	2	7.4
	Israel	1	3.7
School Type (mainstream age ranges)
	Elementary/Primary School (5–11)	15	55.6
	Middle School (11–14)	4	14.8
	Lower Secondary School (11–15)	5	18.5
	High School (14–18)	3	11.1
	Upper Secondary School (15–18)	1	3.7
	Non-Mainstream School	1	3.7
	Unknown	1	3.7
Target child characteristics (for outcomes)
	Familiar (i.e. classmates)	16	59.3
	Unfamiliar (i.e. vignettes)	9	33.3
	Unknown/Not Applicable	3	11.1
Outcomes of interest
	Attitudes	24	88.9
	Behavioural Intentions	10	37
	Behaviours Actual	9	33.3
	Knowledge	7	25.9
Outcome Measures
	Shared Activities Questionnaire	10	37
	Adjective Checklist	9	33.3
	Qualitative interviews/focus groups	8	29.6
	Direct observations of behaviour	7	25.9
	Knowledge questionnaire	5	18.5
	Chedoke-McMaster Attitudes Towards Children with Handicaps	3	11.1
	Other	10	37
Study Design (for outcomes of interest)
	Quantitative	19	70
	Qualitative	8	29.6
	Single-Subject	5	18.5

Participants

Data on a number of peer participants (non-ASD) was reported in 22 of the included studies. These studies collected data on stigma related outcomes for 3,619 peers of children with ASD (range of participants per study 3-576; M = 164.5, SD = 179.64). From these 22 studies, 16 reported on gender: 1,587 males and 1,388 females. The age of participants was difficult to determine. Mean/average age was reported in 12 of the 27 studies (range 6.7–15.3; overall M = 11.13, overall SD = 2.3).

The types of school settings in which the intervention was delivered are summarised in Table 2. For some studies, school type was not reported but could be deduced from the country paired with reported participant grades and ages (reported in brackets in Table 1). As the type of school could be established for most of the included studies (n = 26), it is likely a more accurate reflection of the age range of participants represented in this review. The number of studies conducted in each school setting is elementary/primary schools (approximate ages 5–11), n  = 15; middle schools (approximate age range 11–14), n = 4; lower secondary schools (approximate age range 11–15), n = 5; high schools (approximate age range 14–18), n=3; upper secondary schools (approximate age range 15–18), n = 1; non-mainstream schools, n = 1; and unknown, n = 1.

Participant inclusion criteria for the 17 studies which reported this is summarised in Table 1. Studies assessing stigma towards familiar autistic children regularly recruited those with desirable characteristics such as: likely to engage and comply with the intervention e.g. interest in group work, regular attendance and good social skills (Collet-Klingenburg, Neitzel, and LaBerge 2012; Cook 2017; Ezzamel 2016; Ezzamel and Bond 2017; Gardner et al. 2014; Houston 1998; Hughes et al. 2013; Owen-DeSchryver et al. 2008; Sreckovic 2015; Sreckovic, Hume, and Able 2017); previously demonstrated interactions with the target child (Collet-Klingenburg, Neitzel, and LaBerge 2012); and judged likely to engage with the target child (Gardner et al. 2014; Hughes et al. 2013; Sreckovic 2015; Sreckovic, Hume, and Able 2017). Studies often required participants to be in the same class as the target child (Collet-Klingenburg, Neitzel, and LaBerge 2012; Cook 2017; Ezzamel 2016; Ezzamel and Bond 2017; Hughes et al. 2013).

In contrast, participants in studies assessing stigma without a target child or towards unfamiliar peers with ASD were often those shown to be unknowing about autism (either from never having heard of autism, or an inability to provide a reasonable description/definition) and/or in a school classroom that did not have any autistic children (Campbell et al. 2004, 2005; Fleva 2014, 2015; Morton and Campbell 2008; Scheil, Bowers-Campbell, and Campbell 2017; Silton and Fogel 2012; Swaim 1998; Swaim and Morgan 2001). In one case those classes with higher numbers of children with ASD were allocated to control conditions (Staniland and Byrne 2013).

Target children with ASD

Scheil, Bowers-Campbell, and Campbell (2017) assessed change in knowledge of autism, and was the only study that did not include at least one target child with ASD as part of the intervention delivery and/or assessment. There were 52 familiar target students from 16 studies (59.3%; per study M = 3.31, SD = 3.67). Gender information was available for 44: 38 males and 6 females. Information relating to age was reported in various formats (see Table 1 for details per study), with an estimated mean age of 11.56 (range 5–18) derived from available information. A total of 11 unfamiliar target children were used across nine studies (33.3% of the sample). All were hypothetical children described using vignettes. The gender was unreported for two (Mavropoulou and Sideridis 2014; Silton and Fogel 2012). Most were male (n = 8; Campbell 2007; Campbell et al. 2019; Campbell et al. 2004, 2005; Fleva 2014, 2015; Morton and Campbell 2008; Silton and Fogel 2012; Swaim 1998; Swaim and Morgan 2001), and there was one female (Silton and Fogel 2012). From the data reported, the mean age of unfamiliar target children was 12.71. Two studies (Ranson and Byrne 2014; Staniland and Byrne 2013) had participants complete outcome measures with reference to ‘their peers with autism’ so the level of familiarity could not be categorised. Further descriptions of all target ASD children are reported by a study in Table 1.

Methods

Table 1 summarises the methods used in each of the included studies, and Table 2 provides a summary. Attitudes were assessed in 88.9% of included studies, behavioural intentions in 37%, actual behaviour in a third of all included studies (33.3%), and a measure of knowledge of autism was included in just over a quarter (25.9%). Five studies assessed three outcomes (Campbell 2007; Gus 2000; Mavropoulou and Sideridis 2014; Ranson and Byrne 2014; Staniland and Byrne 2013), and nine reported data for a single outcome: attitudes (Cook 2017; Frederickson, Warren, and Turner 2005; Gardner et al. 2014; James 2011; O’Connor 2016; Reiter and Vitani 2007); actual behaviour (Houston 1998; Owen-DeSchryver et al. 2008); and knowledge (Scheil, Bowers-Campbell, and Campbell 2017). Of the remaining 13 studies, six assessed both attitudes and intended behaviour (Campbell et al. 2004, 2005; Fleva 2014, 2015; Morton and Campbell 2008; Silton and Fogel 2012; Swaim 1998; Swaim and Morgan 2001), six assessed both attitudes and actual behaviour (Collet-Klingenburg, Neitzel, and LaBerge 2012; Ezzamel 2016; Ezzamel and Bond 2017; Hughes et al. 2013; Simpson and Bui 2016; Sreckovic 2015; Sreckovic, Hume, and Able 2017; Whitaker et al. 1998), and one assessed knowledge and attitudes (Campbell et al. 2019).

All but three of the included studies employed a quantitative methodology for at least one outcome of interest (group design, n = 19; and single-subject design, n = 5). Eight studies used a qualitative approach to assess at least one relevant outcome. Within each of these broad methodologies, there was great variability in the approaches taken to evaluate the impact of interventions (all listed in Table 1). In the group designs, a follow-up maintenance assessment was included in seven of the studies, ranging from one-week to 23-weeks.

Reflecting the larger number of quantitative studies included in the current review, standardised pen and paper measures were the most frequently used method of assessment. All were self-report measures. The Shared Activities Questionnaire (Morgan et al. 1996) was used in 10 of the included studies as a measure of participants’ behavioural intentions towards a peer with ASD. The Adjective Checklist (Siperstein 1980; Siperstein and Bak 1977) was the tool favoured to assess attitudes and used in a third of all included studies. The Chedoke-McMaster Attitudes Towards Children with Handicaps (Rosenbaum et al. 1986), or a modified version, was an alternative self-report measure of attitude used in three studies.

Four different knowledge of autism questionnaires were used across five studies: a bespoke measure in Campbell (2007) and an adapted version of this tool (developed by Campbell and Barger 2011) was used in Campbell and colleagues’ 2019 study; Mavropoulou and Sideridis (2014) used the ‘Adapted Knowledge of Autism Questionnaire’ (Ross and Cuskelly 2006); and the bespoke Autism Knowledge Questionnaire, developed and used by Staniland and Byrne (2013), was also used in Ranson and Byrne’s study (2014). Both Gus (2000) and Scheil, Bowers-Campbell, and Campbell (2017) used qualitative methods to explore the impact of the intervention on knowledge of autism.

Focus groups with participants (Collet-Klingenburg, Neitzel, and LaBerge 2012; Ezzamel 2016; Ezzamel and Bond 2017) and participant interviews (Cook 2017; Hughes et al. 2013; Simpson and Bui 2016; Whitaker et al. 1998) were the qualitative methods employed to explore the impact of the intervention on peer attitudes and/or subjective behavioural change towards an autistic peer. Other studies assessing actual behaviour change (rather than behavioural intent; n = 7) used direct observations of participant behaviours as a method of measurement.

Interventions

A summary of the interventions employed by all studies is reported in Table 3 and summarised in Table 4.

Table 3. Intervention characteristics and evaluation results of included studies.

Authors (date) Document Type	Intervention Deliverer	Description	Information Type		Intervention Duration	Results
Campbell (2007) Journal Article	Video vignette(s) Researcher	63-second video of a child engaging in autism-like behaviours and a pamphlet (read aloud) with a picture of the child and one of four types of information about the child: basic (n = 68); descriptive (n = 63); explanatory (n = 42); or descriptive and explanatory combined (n = 60)	Des	✓	1 session	No effect of intervention compared to a no-information control group. Explanatory and combined information resulted in higher post-test scores on knowledge compared to descriptive only; participants unfamiliar with autism reported more favourable attitudes when receiving combined information; Explanatory information resulted in more favourable behavioural intentions compared to descriptive information.
			Exp	✓
			Dir
			ASD	✓
Campbell et al. (2019) Journal Article	Kit for Kids Researcher	An educational poster, peer handouts and lesson plan providing explanatory information about ASD, information that targets misperceptions of ASD, descriptive information about a child with autism, and directive information. Participants viewed a video vignette of a different child displaying autistic-like behaviours and completed measures with reference to this child.	Des	✓	1 session 30–40 min	Compared with controls, there was a significant gain in knowledge. For participants unfamiliar with autism, the intervention improved initial attitudes compared to controls. Significant gains maintained at 1-week follow-up.
			Exp	✓
			Dir	✓
			ASD	✓
Campbell et al. (2004, 2005) Journal Articles	Video vignette(s) Researcher	In classroom groups participants watched a no autism video with descriptive information, and an autism video with either descriptive or combined descriptive and explanatory information. Each video contained a child doing a puzzle with a female adult voice-over delivering the information.	Des	✓	1 session < 3 min	Compared with descriptive information alone, combined information resulted in (1) improved third- and fourth- graders’ but not fifth-graders’ attitudes and (2) improved behavioural intentions for all participants. Females were more responsive to combined information than males for academic behavioural intentions. Rejected children were more responsive to combined information than other sociometric groups for all behavioural intentions except recreational.
			Exp	✓
			Dir
			ASD	✓
Collet-Klingenburg, Neitzel, and LaBerge (2012) Journal Article	Power-PALS peer mediated intervention Teacher/ speech and language therapist/the district pupil services’ director	30-minute PowerPoint adapted from Gray's Sixth Sense II (2002) and an age appropriate book on autism (Keating-Velasco 2007), followed by a group discussion on ways to support and interact with a peer with ASD. Peers were supported by weekly meetings which focused on how participants could better include targets and how they could encourage non-participants to do the same	Des		Weekly sessions: two groups met 11 times, one group met 7 times	Peer initiations per minute towards a target child increased from .175 (SD .177) before intervention to .49 (SD .17) after. Focus group suggested a shift in perspective by participants, with reports of increased patience for deviant behaviour.
			Exp	✓
			Dir	✓
			ASD	✓
Cook (2017) Thesis	Modified LINKS peer mediated intervention Researcher and social worker	Introduction meeting to explain research and for participants to socialise with targets. Orientation meeting for participants to learn about target peers (emphasis on similar interests to establish relationships), make goals, problem-solve methods of peer mediated academic support. Bimonthly supervision/check-in meetings.	Des	✓	8 weeks Orientation meeting and 4 check-ins (30–45 min each), 8 peer contact sessions (55 min each)	Interviews suggest the intervention provided an opportunity for participants to get to know their peers with ASD. One true friendship formed from intervention. Participants highlighted a need for more directive information (instruction). One participant socially rejected their target peer, who reported a lack of confidence/competence.
			Exp
			Dir
			ASD
Ezzamel (2016) and Ezzamel and Bond (2017) Thesis and Journal Article (respectively)	Peer mediated intervention Researcher, teacher and teaching assistant	Two whole-class awareness raising sessions providing descriptive information about the target, explanatory information about disabilities in general, experiential exercises to understand what it might be like to have autism (autism not mentioned), and a problem-solving session to discuss ways the class might help the target. 6 small-group sessions focused on social skill development and ways to interact and support the target. This small group fed back what was learned to the wider class with weekly reports, and to the school in assembly after the final session.	Des	✓	8 weekly sessions	Initiations towards the target child increased from 20.5 at time 1, to 23.5 at time 2, and to 26 at follow-up. Responses increased from 8.5 at time 1, to 18 at time 2, and to 16.5 at follow-up. Qualitative data were suggestive of increases in peer acceptance, understanding of the target child's needs, and increased skills to interact with the target child.
			Exp	✓
			Dir	✓
			ASD
Fleva (2014) Journal Article	PowerPoint vignette(s) Teacher	Participants viewed a PowerPoint picture slideshow of a child without autism. The voice-over was either the friend of the child or teacher providing descriptive information. Participants then viewed a picture slideshow of a child with autism, with information again provided via voiceover by either the child’s friend or teacher. Information was either descriptive, or combined descriptive and explanatory.	Des	✓	1 session 25 min	Non-significant impact of type of information. Participants showed greater behavioural intentions towards the autistic child when introduced by a friend rather than a teacher.
			Exp	✓
			Dir
			ASD	✓
Fleva (2015) Journal Article	PowerPoint vignette(s) Teacher	Participants viewed a PowerPoint presentation of pictures and a recorded vignette describing a child with autism. The voice-over either provided descriptive only information or a combination of descriptive and explanatory information. This was read by a young male described as the target's friend. At the end of the presentation, participants were either asked to imagine the presence of the target child in the class or imagine interacting with the target child.	Des	✓	1 session	Non-significant effect of information type on attitudes. Combined information was significantly better at improving social intentions than descriptive only, but not academic or recreational intentions. Combined information was more effective in the imagined contact condition compared to the presence contact for academic intentions.
			Exp	✓
			Dir
			ASD	✓
Frederickson, Warren, and Turner (2005) Journal Article	Circle of Friends Trained assistant educational psychologist	Delivered as described by Taylor (1996, 1997). The whole-class session was modified to include an explanation of communication and social interaction difficulties in ASD, and how these difficulties influence behaviour. Detailed scripts for the whole class discussion were taken from the Buckinghamshire Educational Psychology Service manual (Turner 1999).	Des	✓	1 whole class discussion 6 circle group meetings	Visual inspection of the data shows increasing acceptance scores across time. This gain was not maintained at follow-up. Rejection scores decreased over the intervention phase and did not return to baseline at follow-up. Decreases coincided with reported decreases in disruptive behaviours in the target child as perceived by participants.
			Exp	✓
			Dir	✓
			ASD	✓
Gardner et al. (2014) Journal Article	Peer Network School staff	Manualised peer network programme that consisted of: an orientation meeting with an introduction, a discussion of shared interests, and discussion about when and how the groups could socialise with specific reference to a deficit in socialisation experienced by the target; social group meetings 1–2 days per week which included some descriptive and directive information.	Des	✓	1–2 days per week for unknown duration	To the statement ‘I consider my partner to be a friend’, 3 participants strongly agreed, 2 agreed, and 1 was neutral; and to the statement ‘My views about students with autism have changed for the better’, 3 participants strongly agreed, 2 agreed, and 1 was neutral.
			Exp
			Dir	✓
			ASD
Gus (2000) Journal Article	Modified Circle of Friends Researcher and form tutor	A whole-class Circle of Friends orientation meeting provided: an explanation of ASD; a description of how autistic people are like non-autistic people; ideas of how people can interact with someone with ASD; and a discussion of the overlap between features of ASD and the difficulties the target presents.	Des	✓	1 session	Qualitative interviews suggest the intervention led to a greater understanding of autism, a shift in attitudes, and a change in behaviour (e.g. being clearer with their language when speaking with the target child). There was maintenance at follow-up. Participants reported being socially inclusive, and had stopped others victimising the target child.
			Exp	✓
			Dir	✓
			ASD	✓
Houston (1998) Thesis	Peer mediated intervention Researcher, RA, teacher and guidance counsellor	Two of the three target children’s classes participated in a lesson from the Sixth Sense (Gray 2002) to increase understanding of: ASD (without mentioning autism); how deficits in social ability can impact behaviour; and the means to support a student peer with ASD. Small groups of three from all classes had social skills training.	Des	✓	1 whole class discussion 2–3 weeks of training	Results suggest no added benefit of the whole-class education lesson in change in actual behaviour towards peers with ASD.
			Exp	✓
			Dir	✓
			ASD
Hughes et al. (2013) Journal Article	Peer mediated intervention Researcher	Participants discussed as a group the target's strengths and weaknesses, how to interact with them based on similarities, and how to set interaction goals for them in class.	Des	✓	1 session 20 min	Peer initiations towards target children increased for all three dyads (from no initiations to interacting 24.9%, 41% and 45.1% of the time on average per session). The interviews suggest participants increased understanding of ASD, and they considered the target child a friend rather than an acquaintance.
			Exp
			Dir	✓
			ASD
James (2011) Thesis	Modified Circle of Friends Researcher, leaning support worker or SENCo	Adapted from Newton, Taylor, and Wilson (1996). A whole class meeting included a discussion of the target’s strengths and weaknesses, information on ASD and how this impacts the target's behaviour (autism as a label was not used), and group problem-solving on how to support the target based on these challenges. A small circle group of 6 members further supported the target to develop appropriate behaviour and had weekly support meetings.	Des		8 weeks 1 whole class discussion 8 circle group meetings (30 min each)	Participants’ ratings of perceived causality of behaviour were unchanged from over time (1.5 SD .548), suggesting no effect of intervention.
			Exp	✓
			Dir	✓
			ASD	✓
Mavropoulou and Sideridis (2014) Journal Article	Bespoke education programme with contact Education intervention not reported	The intervention group (n = 224, 10 classes) participated in an integration programme and had psychoeducation beforehand that involved: reading a children's book about ASD; watching a 5-minute video about ASD from the UK's National Autistic Society; engaging with the ‘understanding friends’ activity from the TEACCH approach; and participating in lessons from the Sixth Sense II (Gray 2002). The control group (n = 251, 12 classes) had no education intervention or structured contact with children with ASD.	Des	✓	Up to 15 weeks 90 min once a week	Compared to controls, the intervention group had more positive attitudes, behavioural intentions and knowledge of ASD. Males showed more positive gain for academic behavioural intentions, and girls for social intentions. Knowledge gain was greater in 4th and 5th grades, and behavioural intentions greater in 5th graders. Effects were maintained at 3 months. Within the intervention group, there were no differences in any dependent variable between contact level groups (high to low contact during integration sessions).
			Exp	✓
			Dir
			ASD	✓
Morton and Campbell (2008) Journal Article	Video vignette(s) Researcher and/or class teacher	In classroom groups participants watched a videotaped vignette providing descriptive information about a target peer with ASD. Groups then received explanatory information from one of five sources: videotape (n = 66); their class teacher (female; n = 58); an actor portraying the target child's mother (n = 57); an actor portraying the target child's father (n = 55); a female actor portraying a ‘doctor’ who worked with children like the target child (n = 60).	Des	✓	1 session < 3 min	Attitudes: females responded more favourably to the videotape than males; fifth-graders responded more favourably to teacher delivered information than third-graders. Behavioural intentions: fifth-graders responded better when information was provided by an extra-familial source versus parent source; fourth-graders were influenced more by teacher delivered message than doctor delivered message; mother yielded more persuasive effects on behavioural intentions for third-graders compared to fifth-graders.
			Exp	✓
			Dir
			ASD	✓
O’Connor (2016) Journal Article	Circle of Friends Not reported	The circle group met weekly to problem-solve ways of helping the target compensate for difficulties in social interactions and how to go about interacting with this child.	Des		3 months 12 weekly sessions (30–40 min each)	Intervention resulted in a positive shift in peer acceptance: 70% increase in participants self-reported willingness to make social initiations and relationships with the target child; 80% more likely to accept deviant behaviour.
			Exp
			Dir	✓
			ASD
Owen-DeSchryver et al. (2008) Journal Article	Bespoke education programme One of two researchers	Second-grade participants read a storybook about autism (Owen-DeSchryver 2002). Fourth-grade students did an adapted Circle of Friends activity. All classes discussed the strengths and weaknesses of the target which were compared to the strengths and weaknesses of all children. Concrete information and strategies were provided that would help participants interact with targets. There was no input for the control group.	Des	✓	3 sessions (30–45 min each)	The impact of intervention on peer responses to target children could not be extracted. There was minimal impact of intervention on peer initiations: an increase of .2 average initiations per minute in the intervention condition and .19 for untrained peers.
			Exp
			Dir	✓
			ASD	✓
Ranson and Byrne (2014) Journal Article	Understanding Our Peers Researcher	Manualised anti-stigma programme using descriptive, explanatory and directive information provided through education, direct and video contact with those with autism, and a web-based learning site for reflection and class discussion. Included some elements from the Sixth Sense II (Gray 2002) and several books and online videos and resources.	Des	✓	8 weeks 8 sessions (50 min each)	The intervention positively influenced knowledge, attitudes and (to a lesser extent) behavioural intentions in the intervention condition. Changes in attitudes and knowledge were maintained at 3 months.
			Exp	✓
			Dir	✓
			ASD	✓
Reiter and Vitani (2007) Journal Article	Mediation Not reported	Class activities targeting intention and mutuality, transfer of knowledge, mediation of meaning, sense of efficacy, participatory behaviour, coping with challenges.	Des		> 19 weeks > 19 sessions (60 min each)	There was a significant difference in before and after scores with participants showing more positive attitudes towards peers with autism at the end of the intervention programme.
			Exp
			Dir	✓
			ASD
Scheil, Bowers-Campbell, and Campbell (2017) Journal Article	Kit for Kids Researchers	Manualised Kit for Kids programme	Des	✓	1 session	From interviews, participants reported the programme supported prior knowledge and experiences of ASD, and also provided novel knowledge of ASD and information about how to interact with someone with ASD. Differences as opposed to similarities between those with and without ASD emerged as a theme. Participants requested more descriptive and directive information.
			Exp	✓
			Dir	✓
			ASD	✓
Silton and Fogel (2012) Journal Article	Video vignette(s) Researcher	Participants viewed one of four videos. All provided descriptive and explanatory information about autism (adapted from the scripts of Campbell et al. 2004) and included actors playing two siblings with autism and their mother. Video 1 (n = 44): combined descriptive and explanatory information only. Video 2 (n = 39): included additional peer strategy component which, using an additional two actors playing neurotypical children, introduced challenges experienced by those with ASD and suggested ways peers could manage these social challenges. Video 3 (n = 31): included additional strengths information section which incorporated information on three strengths/special abilities those with ASD may possess. Video 4 (n = 43): included all four information types.	Des	✓	1 session 22 min	Video 1 (descriptive and explanatory combined) was the only video to have significant change pre- to post- intervention with significant decreases in total behavioural intentions and academic behavioural intentions. Video 3 (descriptive, explanatory and strengths) was also less effective at improving behavioural intentions compared to video 2 (descriptive, explanatory and directive) and video 4 (all types of information).
			Exp	✓
			Dir	✓
			ASD	✓
Simpson and Bui (2016) Journal Article	Peer mediated intervention Researcher and teaching staff	Participants were instructed to stay, read and talk with their peer during literacy lessons. Training was provided on how participants might talk to their peer about a book. Before intervention, the class was provided general information on ASD, read the book Since We're Friends: An autism picture book (Shally and Harrington 2007), and discussed how to have a friend with ASD and that friends can be alike and different.	Des	✓	10 weeks 1 whole class session (20 min) and 20 sessions (15 min each)	Developing friendships was one of three themes that emerged from the interviews with participants who provided examples of how they had become friends with the target children. The impact of the education session on peer initiations and responses towards target children could not be determined (no pre-session measure), but initiations were seen to improve over the directive reading sessions.
			Exp
			Dir	✓
			ASD	✓
Sreckovic (2015), Sreckovic, Hume, and Able (2017) Thesis and Journal Article (respectively)	Peer network Researcher and research assistant	Adapted from the peer network implementation strategies of Carter et al. (2014) which consisted of three components: (a) an orientation meeting which included a reflection on strengths and weaknesses of the targets and participants, and a brainstorm of ways to develop social skills in targets; (b) an introduction meeting where participants and targets were facilitated in a discussion of similar interests and how they might socialise together; and (c) bi-weekly peer network meetings where participants problem-solved barriers and the facilitators encouraged interaction.	Des	✓	5 weeks 10 sessions (25–30 min each)	Visual inspection of the data suggested an upward trend in peer initiations towards target children during intervention with variable maintenance. Data were variable during intervention for peer responses, with an upward trend observed which maintained over time. Target students had limited social engagement before intervention. After, target children were named as friend by the majority of peer network members.
			Exp
			Dir	✓
			ASD
Staniland and Byrne (2013) Journal Article	Understanding Our Peers Researcher	Manualised anti-stigma programme using descriptive, explanatory and directive information provided through education, direct and video contact with those with autism, and a web-based learning site for reflection and class discussion. Included some elements from the Sixth Sense II (Gray 2002) and several books and online videos and resources	Des	✓	6 weeks 6 sessions (50 min each)	The programme significantly improved the knowledge and attitudes towards peers with ASD, but not behavioural intentions. Effects endured at follow-up. Greater engagement with the web-based platform increased knowledge. There were no ‘spill over’ effects of intervention to others in the school.
			Exp	✓
			Dir	✓
			ASD	✓
Swaim (1998), Swaim and Morgan (2001) Thesis and Journal Article (respectively)	Video vignette(s) Researcher and research assistants	In groups of 10, participants watched a video corresponding to one of three conditions: no autism descriptive information (n = 78), autism descriptive information (n = 77), or autism with combined descriptive and explanatory information (n = 78). Each video showed a child doing a puzzle with a female adult voice-over delivering the message.	Des	✓	1 session < 3 min	Combined information about autism did not provide a significant additive benefit to descriptive only information on ratings of either attitudes or behavioural intentions as ascribed to self or others.
			Exp	✓
			Dir
			ASD	✓
Whitaker et al. (1998) Journal Article	Circle of Friends Researchers and school staff	A whole class session provided explanatory information on a target’s strengths and difficulties. Weekly small circle groups focused on collaborative problem-solving for social inclusion.	Des		1 whole class meeting 3–17 circle group meetings 20–30 min each	Qualitative findings suggest the programme increased participants’ understanding of the target students, their tolerance for deviant behaviour, and their social contact with their ASD peers (40/52 self-reporting increased prosocial behaviour). After intervention 3/52 participants referred to the focus child as a friend.
			Exp	✓
			Dir
			ASD

Note: Des = Descriptive. Exp = Explanatory. Dir = Directive. ASD = Autism facts.

Table 4. Summary of intervention characteristics of included studies.

		n	%
Intervention types
	Vignette(s) – video or PowerPoint	7	25.9
	Peer mediated intervention	6	22.2
	Circle of Friends	5	18.5
	Understanding Our Peers	2	7.4
	Kit for Kids	2	7.4
	Peer network	2	7.4
	Unnamed education programme	2	7.4
	Mediation	1	3.7
Information provided
	Descriptive	22	81.5
	Explanatory	19	70.4
	Directive	18	66.7
	ASD facts	18	66.7
Number of pieces of information
	1	4	14.8
	2	3	11.1
	3	13	48.2
	4	7	25.9
Intervention duration
	One session	11	40.7
	Multi-session	16	59.3
Involvement in intervention delivery
	Researchers (including assistants)	20	74.1
	School staff	11	40.7
	Other	3	11.1
	Not reported	3	11.1

Within interventions, descriptive information was the most provided information type (81.5%), followed by explanatory (70.4%). Equal numbers of studies (n = 18) provided participants with directive information and ASD facts. The majority included more than one type of information, with 14.8% using only one. Most used three pieces of information in their intervention (28.2%), and just over a quarter used all four types. This information was delivered mostly by researchers (74.1%), with school staff involved 40.7% of the time.

There were eight different approaches to intervention. Vignettes about hypothetical children with ASD were used in just over a quarter of studies (Campbell 2007; Campbell et al. 2004, 2005; Fleva 2014, 2015; Morton and Campbell 2008; Silton and Fogel 2012; Swaim 1998; Swaim and Morgan 2001). These interventions involved participants receiving information about ASD paired with either a PowerPoint of pictures, or a video of actors portraying an autistic child. All explored the effects of providing descriptive, explanatory and ASD specific information on outcomes, with most comparing conditions to explore the unique impact of information type. Silton and Fogel (2012) also assessed the impact of directive information (i.e. all four types of information). These interventions were all single sessions ranging in length from <3 min to 22 min, and exclusively targeted stigmatisation of ASD using peer education.

Two included studies evaluated the impact of bespoke ASD education programmes on stigma related variables (Mavropoulou and Sideridis 2014; Owen-DeSchryver et al. 2008), while another nine evaluated one of three manualised programmes: ‘Kit for Kids’, ‘Understanding Our Peers’, and ‘Circle of Friends’. The ‘Kit for Kids’ intervention (Organization for Autism Research 2012) is a single-session programme consisting of a lesson plan, an educational poster, and peer handouts that provide information using all four information types. Two studies in this review evaluated this programme (Campbell et al. 2019; Scheil, Bowers-Campbell, and Campbell 2017), while a further two evaluated the adolescent ‘Understanding Our Peers’ (Staniland and Byrne 2013) programme (Ranson and Byrne 2014; Staniland and Byrne 2013). The latter uses a combination of descriptive, explanatory, and directive information with the provision of ASD facts through direct, video and online content.

‘Circle of Friends’ (Pearpoint, Forest, and Snow 1992) was the most evaluated intervention in this review (18.5% of included studies). It is an educational approach that facilitates inclusion through the development of a peer network for an isolated individual. An initial whole class meeting involves providing descriptive and explanatory information about a target child, including a discussion of strengths and challenges. Peers are asked to brainstorm ways of supporting the child in the school environment (directive information). From this meeting, 6–8 volunteers agree to form the target child’s ‘circle’ and meet regularly to problem-solve coping with inappropriate behaviours. Two studies used adapted versions: Frederickson, Warren, and Turner (2005) included ASD descriptive, explanatory and directive information in their whole class meeting along with facts about ASD; and Gus (2000) used a shortened model (only the whole-class meeting) and included all four types of information.

Several interventions described as either Peer Networks (Gardner et al. 2014; Sreckovic 2015; Sreckovic, Hume, and Able 2017) or peer-mediated interventions (with a focus on social behaviours; Collet-Klingenburg, Neitzel, and LaBerge 2012; Ezzamel 2016; Ezzamel and Bond 2017; Houston 1998; Hughes et al. 2013) employed a similar approach to ‘Circle of Friends’. In these studies, participants were provided with ASD education and/or information on a target autistic child (sometimes as a whole class, and sometimes limited to a small group), and a peer group met frequently to problem-solve barriers to inclusion (through training and/or group discussion). Behavioural change in both the peers and the target child were a focus in these studies. In addition, two peer-mediated interventions aimed at improving the academic attainment in the target child, but which also provided stigma related data exploring the impact of the education session on network peers, were included (Cook 2017; Simpson and Bui 2016).

Reiter and Vitani (2007) were alone in this review in evaluating a mediation programme aimed at decreasing burnout in the peers of those with ASD. Of note is that material/activities from Carol Gray’s Sixth Sense (2002) programme were used in five studies (Collet-Klingenburg, Neitzel, and LaBerge 2012; Houston 1998; Mavropoulou and Sideridis 2014; Ranson and Byrne 2014; Staniland and Byrne 2013). Descriptions of interventions for each included study are reported in Table 3.

What is the overall quality of identified studies?

Details of quality appraisals for each study, as judged by each measure, are reported in Table 5 (NICE quality appraisal checklist – qualitative) and Table 6 (NICE quality appraisal checklist – quantitative; summarised in Figure 2), and Figures 3 and 4 (ROBINS-I).

Figure 2. Quantity of studies rated as being of low, unknown and high quality for the assessment of outcomes of interest using the NICE Quality Appraisal Checklist for Qualitative Studies (n = 8) and the NICE Quality Appraisal Checklist for Quantitative Intervention Studies (internal and external validity; n = 24).

Figure 3. Quality appraisal of quantitative outcomes of included studies as judged using the ROBINS-I.

Figure 4. Quantity of studies (n = 24) rated as being of low, moderate, serious and critical risk of bias for the assessment of quantitative outcomes of interest using the ROBINS-I risk of bias tool.

Table 5. Quality appraisals of the qualitative outcomes of included studies as judged using the NICE Quality Appraisal Checklist – Qualitative Studies.

	Theoretical Approach		Study Design	Data Collection	Trustworthiness			Data Analysis						Ethics	Overall assessment
	Is a qualitative approach appropriate?	Is the study clear in what it seeks to do?	How defensible/ rigorous is the research design/ methodology?	How well was the data collection carried out?	Is the role of the researcher clearly described?	Is the context clearly described?	Were the methods reliable?	Is the data analysis sufficiently rigorous?	Is the data ‘rich'?	Is the analysis reliable?	Are the findings convincing?	Are the findings relevant to the aims of the study?	Conclusions	How clear and coherent is the reporting of ethics?
Collet-Klingenburg, Neitzel, and LaBerge (2012)	+	?	-	-	?	-	-	?	-	?	-	+	-	-	-
Cook (2017)	+	+	?	+	+	-	?	+	+	-	+	?	+	+	+
Ezzamel (2016) and Ezzamel and Bond (2017)	+	+	+	+	?	+	+	+	+	+	+	+	+	+	++
Gus (2000)	?	?	?	?	?	-	-	?	-	?	-	+	+	?	-
Hughes et al. (2013)	?	?	-	?	-	+	-	?	-	?	?	+	+	?	-
Scheil, Bowers-Campbell, and Campbell (2017)	+	+	+	+	-	+	+	+	+	+	+	+	+	+	++
Simpson and Bui (2016)	+	+	?	?	?	-	-	?	?	?	+	+	?	?	+
Whitaker et al. (1998)	+	+	-	?	+	-	-	-	?	?	?	+	-	?	+

Note: ? = mixed/unclear/Not sure/Not described/partially relevant. + = Appropriate/adequate/clear. ++ = The study/aspect of the study has been designed or conducted in such a way as to minimise the risk of bias. + = Either the answer is not clear from reporting, or not all potential sources of bias have been assessed for that particular aspect of study design. - = Significant sources of bias may persist.

Two of the studies reporting qualitative outcomes of interest demonstrated congruity between their theoretical approach, study design, methods of data collection, trustworthiness, ethics, and analysis of the data, although failed to provide a clear statement locating the researcher culturally or theoretically in the study (Ezzamel 2016; Ezzamel and Bond 2017; Scheil, Bowers-Campbell, and Campbell 2017). All other studies were rated as unclear or low in quality, mostly owing to unsystematic data collection and/or analysis procedures, and poor study design (see Figure 2).

Regarding quantitative outcomes, across both assessment tools measurement of outcome was identified as a common risk of bias, due (in-part) to the over-reliance on self-report measures and/or unblinded observations of behaviour. Information bias was high for knowledge outcomes, in particular, owing to the use of bespoke measures. Confounding was the greatest source of bias identified by the ROBINS-I (see Figure 4). None of the included studies adequately described/reported both randomisation and allocation concealment, and most failed to control for confounders either through study design or data analysis. This was particularly true of baseline confounding (or selection bias), with the NICE Quality Appraisal Checklist also identifying this as an issue. Only five studies reported recruiting from a representative source population. Most selected participants were unrepresentative of the class/school: recruiting based on desirable characteristics that likely bias the outcome away from the null hypotheses. Nine studies sought participants that would make engagement and compliance with the intervention more likely, and four studies recruited based on a proven, or likely, ability to engage in positive social interactions with the target child(ren). This selection bias impacted negatively on the external validity scores as judged by this tool (see Figure 2).

Across studies, intervention status was generally well defined and various methods were used to enhance consistency of delivery. A member of the research team was involved in intervention delivery 74.1% of the time, and fidelity checklists and manuals for implementation were frequently used. In addition, most of the studies did not omit important outcome data in their reports, and (from those that adequately reported) data were available for all/almost all participants at the study conclusion.

Overall the most methodologically robust studies were the qualitative study conducted by Scheil, Bowers-Campbell, and Campbell (2017), the qualitatively informed results from the mixed-method study reported in Ezzamel (2016) and Ezzamel and Bond (2017), and the quantitative studies of Fleva (2014), Swaim (1998), Swaim and Morgan (2001), Byrne and colleagues (Ranson and Byrne 2014; Staniland and Byrne 2013), and several from Campbell independently or with colleagues (Campbell 2007; Campbell et al. 2004, 2005, 2019; Morton and Campbell 2008).

Is there support for the use of peer education to target ASD stigma?

To ensure rigour in the interpretation of results, assessment of intervention effects was limited to those studies deemed to have good methodological quality (results for all included studies are reported in Table 3 and summarised in Table 4). There were 10 studies meeting these criteria, and comprised those that: evaluated single-session vignette interventions with children unfamiliar with ASD (Campbell 2007; Campbell et al. 2004, 2005; Fleva 2014; Morton and Campbell 2008; Swaim 1998; Swaim and Morgan 2001); explored the utility of the ‘Kit for Kids’ programme (Campbell et al. 2019; Scheil, Bowers-Campbell, and Campbell 2017); and the adolescent ‘Understanding Our Peers’ programme (Ranson and Byrne 2014; Staniland and Byrne 2013); and qualitatively explored the effects of a peer mediated intervention (Ezzamel 2016; Ezzamel and Bond 2017). Interventions effects will be summarised for impact on knowledge, attitudes and behaviours.

Knowledge

An evaluation of the single-session ‘Kit for Kids’ programme was found to be more effective than a control condition at increasing fourth- and fifth-grade students’ knowledge of autism, an effect which maintained over a 1-week period (Campbell et al. 2019). Similar significant differences between control and experimental conditions were found following both a six-session (Staniland and Byrne 2013) and eight-session (Ranson and Byrne 2014) ‘Understanding Our Peers’ programme with eighth-graders. These gains were maintained at a 3-month follow-up. In a qualitative evaluation of ‘Kit for Kids’, participants aged 10–15 described only a general awareness of autism as a type of problem, disorder, or disability before the intervention, and post-interviews could recall specific content from the programme materials about autism-related behavioural difficulties and sensory sensitivities (Scheil, Bowers-Campbell, and Campbell 2017). In contrast, Campbell (2007) found that no information condition (descriptive, explanatory, and combined descriptive and explanatory) was significantly better at improving knowledge of ASD over time compared to a non-information control condition (mean participant age 13.07).

Attitudes

Qualitative data from Ezzamel and Bond’s (2017) peer mediation intervention, with children aged 7/8, suggests attitudinal shifts towards a familiar child with ASD as a consequence of receiving a combination of descriptive, explanatory and directive information. Participants reported increased understanding, and recognition of similarities. Regarding vignette evaluations, Swaim and Morgan (1998, 2001), Campbell (2007) and Fleva (2014) all reported no significant benefit of pairing explanatory information with descriptive information in self-rated attitudes towards a hypothetical autistic child (all using the Adjective Checklist). Campbell (2007) also found no benefit of providing any type of information (descriptive, explanatory, or both combined) over a control no-information condition. Yet, both the six- (Staniland and Byrne 2013) and eight session (Ranson and Byrne 2014) ‘Understanding Our Peers’ programmes were found to significantly improve attitudes (also using the Adjective Checklist) of participants in the intervention condition compared to controls, which maintained at 3-month follow-up.

Campbell (2007) also reported that for participants who had never heard of autism before, attitudes in the descriptive condition were significantly worse than control and combined conditions (which did not differ significantly). Yet, Campbell and colleagues (2019) found that the ‘Kit for Kids’ intervention significantly improved the attitudes of novice participants compared to controls and those who had previously heard of autism (as measured by the Children’s Attitudes Toward Children with Handicaps-Modified tool). Scheil, Bowers-Campbell, and Campbell (2017) qualitative exploration of the same programme on the attitudes of participants’ (who did not have a child with ASD as a classmate) towards a hypothetical ASD peer had mixed findings. While some students described the peer as incapable of doing things, others expressed understanding.

A part-replication of Swaim and Morgan’s (1998, 2001) study with increased statistical power, reported by Campbell et al. (2004, 2005), found that younger participants (those in the third- and fourth-grade) in the combined explanatory and descriptive condition showed improved attitudes towards ASD compared to a descriptive only condition. No such effect was observed for older fifth-grade participants. Morton and Campbell (2008) further found that fifth-graders’ attitudes are significantly improved when the educational intervention is delivered by a teacher as opposed to a parent of a hypothetical child with ASD. Fleva (2014; participants aged 10–16) later found that attitudes are improved when information is delivered by a friend as opposed to a teacher.

Behaviours

Evaluations of the adolescent ‘Understanding Our Peers’ suggest the six-week (Staniland and Byrne 2013) and eight-week (Ranson and Byrne 2014) programmes were ineffective at improving behavioural intentions towards a hypothetical peer over time compared to controls. Scheil, Bowers-Campbell, and Campbell (2017) qualitative evaluation of the ‘Kit for Kids’ programme similarly found that only a few participants could describe action they could take to modify their own behaviours to improve the integration of students with ASD.

Evaluating vignettes, Swaim and Morgan (1998, 2001), Campbell (2007) and Fleva (2014) reported no significant added benefit of explanatory information when paired with descriptive information in self-rated behavioural intentions using the Shared Activities Questionnaire (and in the case of Campbell a control condition). A part-replication of Swaim and Morgan’s (1998, 2001) reported by Campbell et al. (2004, 2005), found participants in the combined explanatory and descriptive condition showed improved behavioural intentions towards a hypothetical child with ASD compared to a descriptive only condition (with no control comparison). Results suggested females were influenced more than males by combined information (particularly for academic intentions; Campbell et al. 2004), as were socially rejected children when compared to other sociometric groups (particularly for social intentions; Campbell et al. 2005).

When in receipt of combined information, Morton and Campbell (2008) found significantly improved behavioural intentions for fifth-graders when the information was delivered by an extra-familial source (i.e. a teacher or ‘doctor’ as opposed to parent of a child with ASD), and for fourth-graders when provided by a mother of a hypothetical child. Although Fleva (2014) found that academic intentions are improved when the teacher delivers a combined educational message as opposed to descriptive only, across both conditions behavioural intentions were higher when the message was delivered by a friend of the hypothetical child as opposed to a teacher.

With regards to actual behaviour change, Ezzamel and Bond’s (2017) focus group evaluation of a peer-mediated intervention reported that although one friendship formed from intervention, participants largely reported difficulty with engaging with their ASD peer. One participant appeared to socially reject the target child, reporting they lacked the confidence/competence to engage. Participants in this study expressed a need for more explicit instruction and information on how to cope with challenging behaviours (i.e. more directive information). A request for additional information to support connection was also made by participants in Scheil, Bowers-Campbell, and Campbell (2017) qualitative study.

Table 6. Quality appraisals of quantitative outcomes of included studies as judged using the NICE Quality Appraisal Checklist – Quantitative Studies.

	Population			Method of Allocation
	Is the source population or source area well described?	Is the eligible population or area representative of the source population or area?	Do the selected participants or areas represent the eligible population or area?	Allocation to intervention (or comparison). How was selection bias minimised?	Were interventions (and comparisons) well described and appropriate?	Was the allocation concealed?	Were participants or investigators blind to exposure and comparison?	Was the exposure to the intervention and comparison adequate?	Was contamination acceptably low?	Were other interventions similar in both groups?	Were all participants accounted for at study conclusion?	Did the setting reflect usual UK practice?	Did the intervention or control comparison reflect usual UK practice?
Campbell (2007)	++	++	+	NR	++	NR	+	++	++	++	++	++	+
Campbell et al. (2019)	++	++	+	+	++	NR	+	++	++	-	++	++	+
Campbell et al. (2005, 2004)	++	++	++	+	++	NR	+	++	++	++	++	++	+
Collet-Klingenburg, Neitzel, and LaBerge (2012)	+	-	-	NA	+	-	-	NR	NA	-	NR	++	++
Ezzamel (2016) and Ezzamel and Bond (2017)	+	NR	-	-	++	-	-	++	NA	NA	++	++	++
Fleva (2014)	+	+	+	+	++	NR	+	++	+	+	++	++	++
Fleva (2015)	-	NR	NR	+	+	NR	+	++	++	+	NR	++	++
Frederickson, Warren, and Turner (2005)	-	NR	NR	NR	++	NR	+	-	NA	NA	NR	++	+
Gardner et al. (2014)	++	-	-	-	+	-	-	+	NA	+	++	++	++
Houston (1998)	-	NR	-	-	++	-	+	++	++	+	++	++	++
Hughes et al. (2013)	++	-	-	-	++	-	-	++	NA	NA	++	++	+
James (2011)	-	-	-	NA	++	NR	-	+	NA	NA	++	++	+
Mavropoulou and Sideridis (2014)	+	NR	++	-	+	++	+	NR	++	+	NR	++	NR
Morton and Campbell (2008)	++	++	++	+	++	NR	+	++	++	++	++	++	+
O’Connor (2016)	-	-	-	NR	+	NR	-	++	NA	NA	NR	++	NR
Owen-DeSchryver et al. (2008)	+	NR	-	-	+	-	+	++	+	++	+	++	+
Ranson and Byrne (2014)	+	-	+	+	++	NR	+	+	++	+	++	++	+
Reiter and Vitani (2007)	+	-	++	NA	-	++	-	NR	NA	NA	++	++	NR
Silton and Fogel (2012)	+	+	NR	+	++	+	+	++	+	++	NR	++	+
Simpson and Bui (2016)	+	NR	NR	NA	++	++	-	++	NA	NA	NR	++	+
Sreckovic (2015) and Sreckovic, Hume, and Able (2017)	++	+	+	-	++	-	-	++	NA	+	++	++	+
Staniland and Byrne (2013)	+	-	+	+	++	NR	+	+	++	+	++	++	+
Swaim (1998) and Swaim and Morgan (2001)	++	++	+	+	++	NR	+	++	++	++	++	++	+
Whitaker et al. (1998)	-	+	NR	NA	+	-	-	NR	NA	NR	NR	++	+

	Outcomes						Analysis					Summary
	Were outcome measures reliable?	Were all outcome measurements complete?	Were all important outcomes assessed?	Were outcomes relevant?	Were there similar follow-up times in exposure and comparison groups?	Was follow-up time meaningful?	Were exposure and comparison groups similar at baseline? If not, were these adjusted?	Was intention to treat (ITT) analysis conducted?	Was the study sufficiently powered to detect an intervention effect (if one exists)?	Were the estimates of effect size given or calculable?	Were the analytical methods appropriate?	Was the precision of intervention effects given or calculable? Were they meaningful?	Are the study results internally valid (i.e. unbiased)?	Are the findings generalisable to the source population (i.e. externally valid)?
Campbell (2007)	+	++	++	+	++	+	++	++	NR	NR	++	+	++	++
Campbell et al. (2019)	+	+	++	+	+	+	+	++	NR	++	++	++	+	++
Campbell et al. (2005) and (2004)	+	++	++	++	+	+	+	++	++	++	++	++	++	++
Collet-Klingenburg, Neitzel, and LaBerge (2012)	-	NR	-	+	+	NR	NA	NR	NR	NA	-	NA	-	+
Ezzamel (2016) and Ezzamel and Bond (2017)	+	++	++	++	NA	++	NA	++	NR	NA	-	NR	+	-
Fleva (2014)	+	++	++	++	++	+	+	++	NR	++	++	++	+	+
Fleva (2015)	+	NR	++	+	++	+	NR	NR	NR	++	++	++	+	-
Frederickson, Warren, and Turner (2005)	+	NR	++	++	NA	++	NA	NR	NR	-	-	-	-	-
Gardner et al. (2014)	-	++	-	++	+	++	NA	++	NR	NA	-	NA	-	-
Houston (1998)	+	++	+	++	++	-	-	++	NA	NA	++	NA	+	-
Hughes et al. (2013)	+	++	++	++	NA	++	NA	++	NA	NA	++	NA	+	+
James (2011)	-	++	+	++	++	++	NA	++	NR	NA	-	+	-	-
Mavropoulou and Sideridis (2014)	+	++	++	++	-	++	NR	++	++	++	+	++	+	+
Morton and Campbell (2008)	+	++	++	++	++	+	+	++	++	++	++	++	++	++
O’Connor (2016)	+	NR	+	++	NA	NR	NA	NR	NR	NR	-	NR	-	-
Owen-DeSchryver et al. (2008)	+	++	+	+	++	++	NR	-	NA	NA	-	NA	-	-
Ranson and Byrne (2014)	+	++	++	++	++	++	-	++	NR	++	++	++	+	+
Reiter and Vitani (2007)	+	++	++	++	NA	++	NA	++	NR	+	++	++	+	-
Silton and Fogel (2012)	+	++	++	++	++	++	++	++	NR	++	-	++	+	-
Simpson and Bui (2016)	+	NR	++	++	NA	++	NA	NR	NA	NA	++	NA	+	-
Sreckovic (2015) and Sreckovic, Hume, and Able (2017)	+	++	++	+	++	++	NA	++	NA	NA	-	NA	+	+
Staniland and Byrne (2013)	+	++	++	++	++	++	-	++	NR	++	++	++	+	+
Swaim (1998) and Swaim and Morgan (2001)	+	++	++	++	+	+	+	++	NR	NR	++	++	+	++
Whitaker et al. (1998)	NR	NR	NR	++	+	-	NA	NR	NR	NA	-	NA	-	-

Note: NR = Not reported. NA = Not applicable. ++=The study/aspect of the study has been designed or conducted in such a way as to minimise the risk of bias. + = Either the answer is not clear from reporting, or not all potential sources of bias have been assessed for that particular aspect of study design. - = Significant sources of bias may persist.

Discussion

This review identified and summarised 27 studies that delivered a peer-directed educational intervention (including all or a combination of descriptive, explanatory, directive or factual information about autism) in a school-setting, and evaluated the impact of this intervention on at least one stigma-related outcome (knowledge of ASD, attitudes towards autistic peers, and actual/intended behaviours towards these peers). Summaries of these studies highlighted a variety in approach to intervention, including peer-mediated, peer network, vignettes, bespoke educational lessons, mediation, and manualised programmes. Methodologies also varied, with a mix of qualitative, single-subject, and group designs represented in included studies. Given the heterogeneity in interventions and methodologies across (and within) studies, a narrative approach to synthesis was used.

With regards to the methodological quality of the included studies, there were few examples of reliable and valid evaluations of intervention. It must be acknowledged that several studies eligible for inclusion in this review did not set out to deliver an ASD de-stigmatisation intervention. Many had a primary focus on change within the autistic individual (e.g. increased social skills), and peer data (although reported) was not a primary outcome. Nevertheless, regardless of the study aim, there were pervasive methodological flaws identified across studies in this review. There was a paucity of randomised control studies, and few employed appropriate control conditions or other means of controlling for confounding variables within and across groups (by design or analysis).

Selection bias was also identified as a pervasive concern. For several studies involving small groups/networks of participants, recruitment was biased towards those with desirable characteristics to increase engagement with the intervention (e.g. compliance, good social skills and academic performance, regular attendance), and not uncommonly a likelihood and/or expressed willingness to develop a relationship with a target child with ASD. In one included study, some participants were replaced during the intervention phase to increase the intervention effect (Owen-DeSchryver et al. 2008). By recruiting participants most likely to display the outcomes of interest, results in these studies were likely biased away from the null hypothesis.

Selection bias was also a concern in larger group designs, which frequently recruited only those unfamiliar with autism and/or from classrooms without autistic peers (i.e. autism novices) to explore the impact of the intervention on the stigmatisation of a hypothetical autistic peer. Preliminary findings from those studies that included participants both with and without previous ASD knowledge and/or contact found these populations differed in how they responded to different types of intervention (Campbell 2007; Campbell et al. 2019). Although the impact of the intervention is important to assess with novice populations, research is also required to evaluate these interventions in integrated/inclusive classrooms. Further, research focused exclusively on hypothetical rather than real peers raises questions of ecological validity. Future research should focus on exploring effects on both familiar and unfamiliar peers with autism.

There is insufficient data to establish the merit of peer education when targeting the stigma experienced by schoolchildren with ASD. To give confidence to the findings of this review, the 10 studies appraised for intervention effects were those rated as having the highest methodological quality from the sample of included studies. Even within this sample, the majority did not include a no-information control group in their design. There was one randomised control trial (randomisation process not reported) which found no effect of vignette intervention (regardless of information type used) on knowledge, attitudes or behavioural intentions (Campbell 2007). Studies that supported short single-session vignette type interventions were post-test only designs, without a control condition, and used self-report measures directly after implementation. A significant difference on a self-report measure directly following intervention is unsurprising, and without a no-information autism condition it is unclear if gains can be attributed to the intervention.

Campbell et al. (2019), Staniland and Byrne’s (2013) and Ranson and Byrne’s (2014) pseudorandomised repeated measures control group designs found a positive effect of intervention for knowledge and attitudes, which maintained at follow-up. There was a non-significant effect of intervention on behavioural intentions towards ASD. These findings provide preliminary evidence that longer, or multi-session education programmes, which provide all four types of information, and delivered through a variety of mediums can reduce ignorance of and prejudice towards peers with ASD. However, these studies also relied on self-report, and the extent to which children’s actual behaviour can be predicted through self-reported behavioural intentions remains to be empirically verified. Future research should address this limitation by including more objective measurements of change, and by conducting high-quality research with actual behaviour as an outcome.

Further qualitative studies could prove invaluable when exploring the processes driving change in stigma. In the current review, Scheil, Bowers-Campbell, and Campbell (2017) qualitative exploration of the ‘Kit for Kids’ intervention found that despite participants reporting attitudinal and behavioural shifts towards an autistic peer, a theme of difference between children with and without autism, or an ‘us’ versus ‘them’ attitude, emerged. This seems detrimental to the aims of the intervention (Scheil, Bowers-Campbell, and Campbell 2017). Similarly, Whitaker et al. (1998) criticised the ‘Circle of Friends’ approach (and similar peer network interventions) as deficit-based, and which places autistic children in the role of helpless support recipient. Thus, although interventions may reduce victimisation, prejudice and discrimination remain.

Three additional limitations of design common across included studies also warrant discussion. Firstly, in most studies a researcher or an expert delivered the educational component of the intervention, raising questions of ecological validity. This is an important variable to consider given that research from this review suggests ‘teacher’ versus ‘other’ delivery impacts intervention outcomes (Fleva 2014; Morton and Campbell 2008). Secondly, those studies judged most methodologically robust neglected to assess intervention effects with both the youngest and oldest schoolchildren. Across all studies, the mean age of participants (from the information reported) was 11 years, with their ASD target peers of a similar age. None of the included studies targeted pre-school/pre-kindergarten children, and infant schoolchildren (kindergarteners). Considering children begin to develop playmate preferences based on a growing awareness of difference at around three years of age (Rubin et al. 2005), interventions aimed at preventing stigmatisation of ASD in younger school-aged children should be developed and evaluated longitudinally for impact as a matter of priority. Finally, although there was a good gender balance of participants across studies, a potential confounding variable was the gender of the target child which was biased towards males. Current estimates of prevalence suggest the male-to-female ratio of autism is close to 3:1 (Loomes, Hull, and Mandy 2017), highlighting a considerable underrepresentation of autistic females in the reported research: only 13.6% and 12.5% of familiar and unfamiliar targets respectively. This limitation of the design is not uncommon in the ASD literature (Stedman et al. 2019), but clearly needs to be addressed in future research.

Limitations and conclusion

Some limitations of this review are worthy of acknowledgement. The review was limited to published evaluations of interventions. There are many examples of known toolkits, media, and campaigns designed to address the stigmatisation of autism in children who have not been empirically investigated, and thus were not considered in the current review. Stigma is also a broad term and the search parameters may not have been sufficient to identify all relevant literature. Related, the heterogeneity of included studies limited synthesis of the data, and a meta-analysis to assess the impact of the intervention was not possible. Furthermore, few studies were returned assessing ignorance (knowledge) as an outcome, and no studies rated as high in methodological rigour quantitively assessed actual behaviour change in participants. This limited interpretation.

Despite these limitations, this review provides some initial insight into the breadth of approaches used to target the stigmatisation of autism in school-aged peers. Limitations such as biased participant selection, the absence of control comparisons, a reliance on post-test only and self-report measurement, and variability in approach to intervention all make it difficult to draw firm conclusions. A critical appraisal of those 10 studies identified as the most robust in design provide some preliminary evidence for both benefit and feasibility of targeting ignorance and prejudice in school-age children, but as of yet not discrimination.

No one approach to intervention for this population can be determined as most effective, but results did highlight some positive findings for change in knowledge and attitudes when using longer programmes, combining all four types of information, delivered using a variety of mediums (i.e. ‘Kit for Kids’ and ‘Understanding Our Peers’). The results of this review also tentatively suggest the effectiveness of intervention could be improved by adapting delivery and content for different audiences (e.g. the person delivering the message). There is a need for further high-quality research to establish the utility of peer education de-stigmatising interventions for the inclusion of autistic classmates, and to determine what components of the intervention are most effective at improving stigma across domains. The results of the current review will hopefully inform these endeavours and future practice.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

S. Morris

Dr. S. Morris is a clinical psychologist working in child and adolescent mental health services.

G. O’Reilly

Dr. G. O'Reilly is the course director of the doctorate in clinical psychology programme University College Dublin.

J. Nayyar

Ms. J. Nayyar is a Ph.D. candidate in Psychology, University College Dublin.

Notes

1 The authors acknowledge the current debate in the literature regarding the language of describing individuals who meet criteria for autism spectrum disorder and how language may accentuate stigma (e.g. Gernsbacher 2017; Kenny et al. 2016; Robison 2019). This paper alternates between the use of person-first (e.g. person with autism) and identity-first (e.g. autistic person) language for those with and without autism, and between the terms autism and autism spectrum disorder.