Introduction

The prevalence rate for autism spectrum disorder (ASD) is approximately one in 59 (1.7%; Centers for Disease Control and Prevention 2018; Randall et al. 2016) and the rate of students who met the criteria for ASD at one college in the US was also found to be between 0.7% and 1.9% (White et al. 2011). Graduation rates for post-secondary students with ASD, however, are low. Newman et al. (2011) found only 39% of students with ASD in the US had graduated from a post-secondary educational institution (including 2-year community college, vocational, business, technical school, or 4-year college) compared with 52% for the general population, and 41% for all students with a disability. Furthermore, in Australia, only 19% of adults with ASD have a post-secondary qualification (certificate, diploma, bachelor’s degree or above) compared with 50% for all adults with a disability and 59% for adults without disabilities (Australian Bureau of Statistics 2012).

The quality of life of adults with ASD is poor when compared to normative rates of competitive employment, independent living, and time socializing with friends (Bishop-Fitzpatrick et al. 2016). Among the general population, a post-secondary qualification typically leads to better quality of life including increased rates of employment and higher wages (Baldwin et al. 2014; Ma et al. 2016). Specifically, in Australia the unemployment rate for persons with ASD falls from 69% for those who finish school in year 10, to 40% for those who complete year 12, and to 34% for those with a bachelor’s degree or higher (Aspect 2013).

Students with ASD may have strengths that assist them in post-secondary education. For example, many reported having a strong memory, high level technical skills, and an intense interest in the subject they are studying (Anderson et al. 2018). The core features of ASD, namely poor communication and social skills, and restricted interests and activities (American Psychiatric Association [APA] 2013), however, place students with ASD at risk for significant academic and non-academic challenges (Jansen et al. 2017). Indeed, a recent review of the literature (Anderson et al. 2017) revealed some of the issues most commonly identified by post-secondary students with ASD. Academic difficulties included problems with understanding abstract or ambiguous concepts, poor planning skills, a tendency towards procrastination, and struggles with group work, presentations, and social skills during class. The most common non-academic difficulties identified included difficulties with socialization, anxiety, depression, sensory sensitivities, and everyday living tasks. Communication, organizational and time management skills were noted to impact both academic and non-academic demands. In addition, students with ASD have reported struggling with the high volume of work and unpredictability of university life (Van Hees et al. 2015), and many are reluctant to disclose to disability services, precluding or delaying potentially beneficial supports (Cai and Richdale 2016).

The characteristics of post-secondary education and educational settings may also present difficulties. Non-inclusive teaching or evaluation methods may prevent students from demonstrating their knowledge, while reduced structure, larger class sizes (compared with high school), and crowds may be a distraction and/or provoke anxiety (Jansen et al. 2017). Some students have indicated that they felt compelled to avoid certain areas on campus which increased their sense of isolation and loneliness (Madriaga 2010). Finally, students may experience long delays waiting for support due to limited resources (Anderson et al. 2018; Jansen et al. 2017) allowing potentially avoidable problems to develop.

Post-secondary educational institutions have increasingly recognized the needs of students with disabilities, including those with ASD, and most offer a range of academic and non-academic supports such as exam accommodations, extensions for assignments, scribes, tutors, mentors, and counselling (Jansen et al. 2017; Sarrett 2018). In addition, some post-secondary educational institutions offer specialist supports for students with ASD (Barnhill 2016; Collegechoice 2017; Sarrett 2018). Until recently, there has been little empirical research into the effectiveness of these interventions (Ashbaugh et al. 2017; Jansen et al. 2017; White et al. 2016).

Kuder and Accardo (2017) have provided (to date) the only systematic literature review of the empirical research on supports used to assist post-secondary students with ASD. They reported the outcomes of eight studies and concluded that the unique needs of post-secondary students with ASD suggest that specialist autism programs were more likely to meet their diverse needs. They also identified important directions for future research. Their review, however, had several limitations in both search procedures and analysis. Their search included only a limited number of keywords, so they may have missed some relevant articles, and only eight studies with 147 participants were located. In addition, the review was limited to studies published in peer reviewed journals, so publication bias may have been an issue (Schlosser et al. 2007). Further, they did not report interrater reliability on article selection, coding or analysis, and there was no assessment of the quality of the studies reviewed which limited the strength of their conclusions. The purpose of the current systematic review is to identify and analyze the studies that report the empirical outcomes of interventions used to support post-secondary students with ASD, with a view to informing staff, academics, and students with ASD, on their effectiveness and to provide direction for future research. The research questions were:

  1. 1.

    What were the characteristics of study participants?

  2. 2.

    What were the study designs and what were their quality?

  3. 3.

    What dependent variables were examined?

  4. 4.

    What independent variables were examined?

  5. 5.

    How effective were the interventions?

Methods

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement protocol for conducting systematic literature reviews (and meta-analyses) guided the methodology and reporting of this review (see Fig. 1). The PRISMA statement, that includes a 27-item checklist and four phase flow diagram, was devised to improve the quality and consistency of systematic review reporting. It includes an explanation and elaboration document that provides the meaning and rationale (with examples) for each checklist item (Moher et al. 2009).

Fig. 1
figure 1

PRISMA flowchart. Study selection process

Full size image

A search for empirical studies of interventions used to assist post-secondary students with ASD was made using six electronic databases, namely A + Education, Education Research Complete, ERIC, ProQuest Dissertations and Theses Global, PsycINFO, and PubMed, in August 2017. The abstracts and titles were searched using the following search terms: (autis* OR asperge*) AND (college OR university OR “higher education” OR tertiary OR post-secondary OR postsecondary OR “further education” OR TAFE), AND (program* OR treatment OR training OR therapy OR model* OR service OR intervention OR “best practice”) NOT (child* OR “intellectual disability”).

Following the database search, a hand search of the article titles from nine autism specific journals between 2007 and 2017 (Advances in Autism, Autism, Autism Research and Treatment, Education and Training in Autism and Developmental Disabilities, Focus on Autism and Other Developmental Disabilities, Good Autism Practice, Journal of Autism and Developmental Disorders, Review Journal of Autism and Developmental Disorders, and Research in Autism Spectrum Disorders) was conducted. In addition, two journals that specialized in post-secondary education (The Journal of Postsecondary Education and Disability and The Journal of College Student Development) between 2007 and 2017, were also searched. To reduce publication bias no restrictions were placed on publication source or date.

The abstract and titles of the identified records were then screened as to whether: (a) the participants were postsecondary students with an autism spectrum disorder; (b) the participants had not been identified as having a co-morbid intellectual disability; (c) the researchers analyzed a specific intervention or program designed to address the needs of post-secondary students with ASD; (d) the intervention was delivered by a post-secondary institution or in collaboration with a post-secondary institution; (e) quantitative student outcome data were reported; and (f) the study was published in English.

After the initial screening of abstracts and titles, the full text articles which remained were screened against the same eligibility criteria. If some of the participants did not meet the eligibility criteria, the study was included if it was possible to extract the data relating to the eligible participants without confounding the interpretation of the study. Articles that described and assessed transition to postsecondary education programs were only accepted if the support was provided by a post-secondary educational institution. Finally, due to the difficulty and expense in obtaining a formal diagnosis which may prevent some students from obtaining formal proof (Beardon and Edmonds 2007), studies were included even where the participants only provided a self-diagnosis of ASD.

The reference lists from the included studies were also searched, and Google Scholar was searched for articles that cited any of the articles selected for inclusion. Finally, all newly identified articles were also screened against the eligibility criteria, and their reference lists and Google Scholar were searched for additional citations.

Data Extraction and Coding

To assist the analysis, all articles were independently coded for participant demographic features (age, country, diagnosis, comorbid conditions, enrolment status); study design; and dependent and independent variables. Detailed information was provided on all data extraction sheets (available from the authors on request). As the purpose of the review was to examine empirical evidence, only quantitative data were extracted.

Study Quality Assessment

The general research classifications suggested by Campbell and Stanley (1963) were used. Randomized controlled trials (RCT) were classified as true experimental designs. Non-randomized comparison group designs with pre-test and post-test as well as time series studies were classified as quasi-experimental. Non-randomized comparison group designs with post-test only, single group pre-post and post-test only designs were classified as pre-experimental. All true experimental and quasi-experimental group (pre-post non-randomized comparison groups) and quasi-experimental single case studies were assessed for quality. Pre-experimental studies were not assessed for quality because their outcomes are fundamentally uninterpretable with respect to causation (Campbell and Stanley 1963).

RCTs were assessed with the checklist devised by Leong, Carter, and Stephenson (2015) and it included: random assignment, blinding of outcome assessment, participant attrition, and pre-test equivalence. A score of zero to two points was awarded for each item.

All quasi-experimental time series studies were assessed for quality with a checklist devised by Preston and Carter (2009, Table 1). The Preston and Carter checklist was an adapted version of the Horner et al. (2005) Quality Indicators Within Single-Subject Research checklist (p. 174), and it assessed: diagnostic information; setting; participant selection criteria and processing; dependent and independent variables; baseline procedures; experimental control; internal validity; external validity; and social validity. To score, each item is awarded up to 10 points, except for external validity which is limited to five points due to single subject designs intrinsically having limited external validity. The maximum score that can be awarded with the Preston and Carter checklist is 65 quality points, with higher scores indicating better study quality.

Table 1 Quality assessment multiple baseline design studies
Full size table

Inter-Rater Reliability

The first author, and either the second, or third independently completed the initial screening of all abstracts, titles and full text articles. Data extraction, coding and quality assessment were independently assessed for all studies by the first author and one of the other three authors. Interrater reliability was calculated as the number of agreements divided by the total number of records/articles and multiplied by 100. Interrater reliability for the initial abstract and title screening was 98%, the full text screening was 87%, coding of the data was 93%, and assessment of article quality was 95%. All differences were resolved by consensus.

Results

The database search identified 364 records after excluding duplicates and five studies were located incidentally, resulting in 369 records (Fig. 1). The hand search of the titles from the nine autism specific journals and the two post-secondary educational journals revealed no further studies. The abstracts and titles of the identified records were then screened against the eligibility criteria and 55 records remained. These studies were read in entirety and then screened against the same eligibility criteria leaving 23 studies. The reference lists of the selected articles were also searched for relevant articles, but no additional studies were located. Finally, Google Scholar was searched for articles that had cited the chosen articles, and one additional article was located bringing the total number of empirical articles reviewed to 24. Most studies were reported in a peer reviewed journal (n = 20; 83%) while the remaining studies (n = 4; 17%) were dissertations. Also, most studies were conducted in the USA (n = 16; 67%), with the balance coming from Australia (n = 2; 8%); the UK (n = 2; 8%); Canada (n = 1; 4%); Ireland (n = 1; 4%); Israel (n = 1; 4%); and Japan (n = 1; 4%).

Participants

There was a total of 291 participants, and their ages ranged from 16 to 28 years (m = 20.4 years). Gender was specified in all but three studies and 83% of participants were male (n = 196) and 17% female (n = 39). Six studies recorded the enrolment status of the participants (91% in those studies were full time). Half the studies reported where the participants lived, and those studies revealed 63% (n = 80) lived on campus, 24% (n = 30) at home, and 13% (n = 17) independently. The severity of the ASD symptoms were only reported in two studies (Ashbaugh et al. 2017; Pugliese and White 2014).

Race/ethnicity or socio-economic details were recorded in 13 studies (12 US studies plus one UK study). In those studies, participants were classified as Caucasian (114, 87%), Hispanic (8, 6%), African American/Black (5, 4%), Asian (3, 2%), and other (1). Socio-economic data was recorded in only one study (Kelly 2008).

Comorbid conditions of participants were described in 10 studies (42%, Table 2). The most common condition was anxiety which was reported by at least some of the participants in all studies that reported comorbid conditions. Other conditions mentioned included depression (n = 7; 29%), attention deficit hyperactive disorder (n = 5; 21%), obsessive–compulsive disorder (OCD; n = 4), agoraphobia, dyscalculia, dysgraphia, dyspraxia, dyslexia, dysthymia, oppositional defiant disorder (ODD), panic disorder, specific phobia, and Tourette syndrome (all n = 1).

Table 2 Study characteristics and findings from quantitative data
Full size table

Study Designs

A diverse range of research designs were used and are summarized in Table 2. There was one “true” experimental (RCT), six quasi-experimental time series, and 17 pre-experimental designs.

Study Quality

Most of the studies reviewed were pre-experimental (n = 17; 71%) and these were not assessed for quality. White et al. (2016) provided the only randomly controlled trial (RCT) and it was reported as a feasibility pilot with only eight participants. It used two alternative treatment groups but no non-treatment group. The study scored four out of a possible eight quality points on the RCT quality checklist as it relied extensively on self-reports, and they did not specify how the simple randomization of participants was conducted. In addition, there was no adjustment for pre-test differences.

There were six quasi-experimental studies, and all used a multiple baseline design (Tables 1, 2). These studies all scored at least 45.8 out of 65 (71%) on the quality assessment checklist (Table 1). Three unequivocal demonstrations of experimental control were only reported in one-third (n = 2) of the studies, and an overt measure for fidelity of implementation of the intervention was only reported in half (n = 3) of the studies. In addition, only half demonstrated acceptable baseline stability. The diagnostic instrument for ASD was only identified in two studies (33%), and a rating of autism severity was only provided in one study (17%). Finally, a formal measure of practicality or cost effectiveness of the intervention was only provided in one study (17%).

Dependent Variables

The dependent variables are summarized in Table 3. In nine studies (37.5%) all dependent variables were non-academic, while in four studies (16.7%) they were all academic, and 10 studies (41.6%) included both academic and non-academic dependent variables. Generalization and/or maintenance was assessed in only eight studies and student satisfaction with the intervention was reported in 13 studies (54%; Table 3).

Table 3 Dependent variables for empirical findings
Full size table

The dependent variables covered a wide range of non-academic difficulties including social skills (n = 11, 46%), mental health (n = 7, 29%), executive function (n = 5, 21%), self-efficacy (n = 2, 8%), communication confidence (n = 1, 4%), future orientation (n = 1, 4%), motivation (n = 1, 4%), self-advocacy (n = 1, 4%), and self-determination (n = 1, 4%). However, sensory sensitivities were not measured in any study. The most common academic dependent variable was change in grade point average (n = 5, 21%). Other academic dependent variables included: number of courses or assessments passed (n = 3, 12.5%); graduation rates (n = 1, 4%); retention rates (n = 1, 4%); self-regulated learning; (n = 1, 4%); and writing quality (n = 1, 4%).

Independent Variables

A broad range of interventions was revealed (Table 4). The interventions were designed to help students develop skills that would enable them to cope with the demands of university life, and 10 also provided in-situ assistance (e.g., mentors attend social activities with their mentee). Interventions included social skills instruction (n = 11, 46%), specialist autism programs (n = 7, 29%), support groups (n = 7, 29%), cognitive behavioral therapy (n = 4, 17%), occupational therapy (n = 4, 17%), academic writing skills (n = 1, 4%), and an academic learning strategy (n = 1, 4%). These interventions will be described separately though some studies included several independent variables.

Table 4 Independent variables
Full size table

Technology

Technology was integral to the intervention in seven studies (29%). Video self-modeling was used to teach social skills by Mason et al. (2012) and Pierce (2013), while White et al. (2016) used a brain-computer interface with virtual reality avatar to teach facial emotion interpretation and to practice social interactions. Jackson et al. (2018) described the use of Microsoft PowerPoint Smart Art Graphics™ software (Microsoft Office 2018) to create graphic organizers that were used to teach writing skills, and the participants in Quinn et al. (2014) were encouraged to used texting or E-mail (or phone) to communicate with their occupational therapist. Kelly (2008) described a goal planning program that used multi-media technology, while McCoy et al. (2014) used a biofeedback intervention that was designed to self-monitor and thereby reduce anxiety.

Programs that Used Students Without Disabilities to Implement an Intervention

Programs that used students without disabilities to assist with the implementation of an intervention were described in 17 studies (67%; Table 4). Student roles included mentors, orientation leaders, participation assistants, student ambassadors, and therapists (supervised doctoral clinical psychology students; White et al. 2016). In all studies that used students without disabilities to support an intervention, the students were experienced and/or trained, and most were studying in relevant fields (e.g., psychology, occupational therapy). Most were also supervised by qualified staff or faculty. Student support programs or interventions were conducted between half and 2 h per week, but more typically for 1 h per week.

The students without disabilities assisted with a range of concerns. The most common was socialization or social skills (n = 13, 81%) though the independent variable specifically addressed social skills in only two studies (Ashbaugh et al. 2017; Koegel et al. 2013). Other concerns supported by student helpers included time management and/or organizational skills (n = 13, 81%), academic concerns (n = 11, 69%), mental health or stress (n = 8, 50%), problem solving (n = 7, 44%), and self-advocacy (n = 6, 38%; Tables 2, 4). In addition, self-efficacy, future orientation, relationship building, personal communication, daily living skills, and positive coping skills were supported in the intervention described by Pearlman-Avnion (2016).

Social Skills and Socialization Interventions

Social skills and/or socialization were a component of 11 studies (46%; Table 2) but were the main focus of only six studies (25%). Ashbaugh et al. (2017) and Koegel et al. (2013) described a structured social planning intervention that was designed to increase socialization and reduce feelings of loneliness or isolation; Mason et al. (2012) and Pierce (2013) both studied the effects of a video modeling social skills intervention; Pugliese and White (2014) described a group social skills problem-solving therapy; and White et al. (2016) compared a psychosocial intervention (CLS) with a virtual reality social skills and facial emotion recognition intervention (BCI-ASD).

Specialized Autism Programs

There were seven (29%) specialist autism programs (Table 2) that included support for advocacy, anxiety, social skills, and campus orientation. Program lengths ranged from a few days (n = 2) to a full academic year or beyond (n = 5).

Support Groups

Hillier et al. (2018) examined a psychosocial support group intervention which included weekly supervised discussions on time and stress management, group work, and social communication. Support groups were also used to supplement individualized programs, including mentoring programs (e.g., Ames et al. 2016; Pearlman-Avnion 2016; Rando et al. 2016; Siew et al. 2017), individualized cognitive behavioral therapy (CBT; White et al. 2016), and group CBT (Furuhashi 2017). The supplementary support groups provided opportunities to socialize with like-minded peers (e.g., Ames et al. 2016) and to practice newly learnt skills (e.g., White et al. 2016). In addition, instruction was given on relevant topics such as reducing exam stress (e.g., Ames et al. 2016), managing group work (Hillier et al. 2018), and coping with daily living (Pearlman-Avnion 2016).

Cognitive Behavioral Therapy (CBT)

CBT was provided in four studies (17%) and was used to address a range of issues, including: stress, anxiety, and depression (n = 2); social problem-solving skills (n = 1); emotion regulation, time management and social skills (n = 1). In addition, Gun et al. (2017) described a behavioral skills training program. The researchers used observation with immediate feedback to teach the social pragmatic and executive function skills needed to meet pre-school practicum requirements. Furuhashi (2017) and Pugliese and White (2014) provided CBT in a group setting while the remaining interventions were conducted individually.

Occupational Therapy Support Programs

An occupational therapy program was evaluated in two studies (Quinn et al. 2014; Schindler and Cajiga 2015). Schindler and Cajiga (2015) described a transition program that used master’s level occupational therapy student mentors to assist the transition of first year college students with ASD, while the program described by Quinn et al. (2014) was staffed by qualified occupational therapists and was available for the duration of the participant’s course. Both programs provided one-on-one support for individualized goals including time management, organizational skills, academic skills (e.g., study and writing skills), social skills, emotional difficulties, and everyday living skills (e.g., healthy living, residential life, leisure skills).

Academic Supports

Although there was an academic component in 14 studies (62.5%; e.g., in Ames et al. 2016 the mentors assisted with coursework), the independent variable specifically addressed academic skills in only two studies (Table 2). Jackson et al. (2018) described an academic writing learning strategy, while Ness (2013) described a self-regulated learning strategy.

Intervention Characteristics: Training, Intensity and Delivery Formats

Intervention personnel included faculty, qualified staff, and peers (Table 4). Session length ranged from 10 min for self-directed neurofeedback (McCoy et al. 2014) to 2 h for a transition program (Schindler and Cajiga 2015) and a psychosocial intervention (White et al. 2016) but the mode for intervention length was between 30 and 60 min (n = 14; 77%). The number of sessions per week ranged from one to five (Table 4) with the more intensive programs mostly being staffed with peers.

The interventions were conducted in various formats. Half (n = 12, 50%) used one-on-one, a quarter (n = 6, 25%) group instruction, and seven (n = 29%) included both individual and group activities (Table 2). All interventions in this review included a component that was intended to assist the participants to develop skills that would help them cope with the demands of university life (e.g., CBT, social skills training), but only ten studies included an intervention component where participants were assisted in situ (e.g., the mentor attended a social activity with the mentee or assisted the mentee in class etc.).”

The frequency of sessions per week were one (n = 11; 61%), one or two (n = 4; 22%), and three to five sessions per week (n = 3; 17%). Study length ranged from a three-day summer transition program (Lambe 2015), to a full academic year (n = 8, 33%), but the mode was 9–14 weeks (n = 11; 46%).

Outcomes

Given the limitations in inferring causation from pre-experimental studies, those studies will be specifically identified when presenting the results.

Satisfaction

Participant service satisfaction (Table 3) was reported in only one “true” experimental (RCT; White et al. 2016) and two quasi-experimental (multiple baseline) studies (Jackson et al. 2018; Kelly 2008). All the participants in White et al. (2016) reported that the program had been helpful though no behavioral changes were observed. In contrast, the participant satisfaction findings reported in Kelly (2008) were inconclusive and participants were also ambivalent about the merits of the intervention. All three participants in Jackson et al. (2018) indicated that they would continue to use the writing strategy most of the time. Most participants from 11 pre-experimental studies reported that the intervention was acceptable and/or helpful (Tables 2, 3).

Academic Achievement

There were two studies with a specific academic independent variable but only Jackson et al. (2018) used a (quasi) experimental design. Each participant of the writing skills intervention described by Jackson et al. improved their writing quality, passed their writing course, and generalized their newly acquired skills to a content specific writing task. In addition, all participants indicated that they continued to use some aspect of the strategy at follow-up. In contrast, only one of three participants in the pre-experimental study described by Ness (2013) increased their overall grade point average (though two participants reported improvements with some courses).

Academic achievement was an apparent collateral benefit following two quasi-experimental interventions in which non-academic concerns were targeted. Grade point average was found to have improved following a structured social planning intervention in both Ashbaugh et al. (2017) and Koegel et al. (2013). Gains in academics were also reported in some pre-experimental studies that included general support for academic and non-academics, including the transition programs of Rando et al. (2016), Shmulsky et al. (2015), and Siew et al. (2017); the support group in Hillier et al. (2017); the peer mentoring programs of McLeod and Harrison (2013), and Schindler and Cajiga (2015); and the occupational therapy support program of Quinn et al. (2014).

Social Skills and Socialization

Social skills and socialization were taught in various ways. Koegel et al. (2013) reported clinically significant gains following a 22-week structured social planning and social skills intervention, while a similar 10-week intervention described by Ashbaugh et al. (2017) found only two (66%) participants made clinically significant gains (though the remainder reported some improvement).

Mason et al. (2012) and Pierce (2013) used a multiple baseline design to analyze the effectiveness of a video modeling social skills intervention. Although all participants were found to have improved their social skills, the improvement for half the participants was small to moderate. Further, White et al. (2016) conducted a small randomly controlled trial of a psychosocial intervention (CLS) and a virtual reality social skills and facial emotion recognition program (BCI-ASD) and found no clinically significant change in personal-emotional or social adjustment from either intervention.

Social skills and/or socialization was a component of 11 pre-experimental studies. (Table 2). In most of these studies the participants reported high levels of satisfaction with their programs with respect to social skills instruction and socialization.

Mental Health

White et al. (2016) investigated the effectiveness of a group-based CBT psychosocial intervention. Although all the participants in the psychosocial group (N = 4) indicated that the program was helpful, no statistically significant improvements were noted in self-regulation of emotions, self-restraint or motivation for two participants (50%). Of six pre-experimental studies that included a mental health component, all authors reported a benefit for some participants, but not all dependent variables recorded behavioral change and in some studies the findings were equivocal.

Transition

The results from a transition program were reported in six studies. Kelly (2008) used a quasi-experimental (multiple baseline) design to assess an 8-week goal planning component of a year-long private residential transition program for students with Asperger syndrome. The findings were inconclusive with respect to self-determination and goal planning, and participants were only mildly supportive about recommending the program.

A transition intervention that spanned the academic year was examined in five pre-experimental studies. Improved academic performance and social and communication skills were described, and stress and anxiety were reported to have reduced (Table 2).

Discussion

Many post-secondary educational institutions provide supports for students with disabilities, and specialist programs for students with ASD are becoming more prevalent, but there is limited research on their effectiveness (Barnhill 2016). The purpose of this paper was to review the studies that reported empirical evidence of interventions that have been used to support post-secondary students with ASD and to assess their effectiveness.

There were 24 studies (291 participants) reviewed (Table 2) which considerably extended the only prior review of eight studies (147 participants) by Kuder and Accardo (2017). Three of the articles reviewed by Kuder and Accardo were excluded from our analysis as they did not meet our inclusion criteria: Jansen et al. (2017) because they reported the findings of a survey of experiences and support use as opposed to an intervention; Koegel et al. (2016), because the purpose of their study was to investigate the experiences of adults with ASD rather than post-secondary students with ASD; and Weiss and Rohland (2015), because no quantitative data were reported. The current review included an additional 19 studies and 197 participants but, nevertheless, the empirical research on post-secondary interventions for students with ASD remains limited.

Participants

A notable feature of the reviewed studies was the constricted reporting of participant characteristics. Of the studies that reported comorbidities, anxiety and depression were notably prevalent, supporting the conclusion of Gelbar et al. (2015) that these issues are a concern for most students with ASD. However, the severity of ASD symptoms was reported in only two studies and comorbid conditions were only reported in ten studies. Similarly, limited demographic details were provided. Only a few studies specified the number of courses studied, confirmed the number of hours participants attended university, or the number of accommodations they were provided. Given the degree of variation in symptomatology possible in ASD, reporting of participant features may be important to consumers who wish to know the characteristics of students for whom the intervention was successful. Consequently, while participant deficits pertaining to the dependent variable were usually specified, the limited details of ASD severity and other comorbid conditions and demographic features reduced the external validity of the reviewed interventions.

Study Design and Quality

The overall quality of the studies was poor. There was only one very small true experimental study and no group quasi-experimental studies. A quarter of the studies (n = 6) were high quality single subject quasi-experimental studies (time series multiple baseline), and three unequivocal demonstrations of control were provided in two of these studies. Moreover, two-thirds of the studies (n = 17, 71%) used pre-experimental designs that are fundamentally uninterpretable regarding causation. Nevertheless, such studies often provided information on the feasibility of interventions and offered preliminary indications of potential direction for more rigorous research. There was a lack of generalization and/or maintenance data in most studies and this may weaken our ability to gauge effectiveness. In some studies, there was very scant empirical data (e.g., Longtin 2014; Rando et al. 2016), and many of the studies were less than 12 weeks in duration (Table 4). Thus, caution is needed when interpreting the results.

Dependent Variables

Non-academic dependent variables were more prevalent than academic dependent variables which may reflect the balance of concerns experienced by students with ASD at post-secondary education (Gelbar et al. 2015; Jansen et al. 2017; White et al. 2016). It has been widely recognized that non-academic difficulties such as mental health issues and social skills difficulties can have a negative impact on academic performance (Koegel et al. 2013; Kuder and Accardo 2017; Siew et al. 2017; Ward and Webster 2017). Class participation, group work and presentations require communication, social, and negotiation skills and can blur the dichotomy of academic and non-academic skills and support requirements in educational setting. Thus, given the main purpose of post-secondary education for most students is to gain a qualification, it may be interesting to investigate whether non-academic interventions (e.g., social skill or mental health supports) have an impact on academic achievement. While some studies did include both academic and non-academic dependent variables (e.g., Siew et al. 2017), 10 studies (42%; Table 3) had no academic dependent variable and only half the studies included a measure addressing course results (e.g., change in grade point average, course grade, retention rate). Given the high non-completion rate of students with ASD (61%; Newman et al. 2011) this information may be particularly relevant.

An evaluation of participant service delivery satisfaction increases social validity and was included in more than half the studies. In addition, both objective and self-reports were included in two-thirds of studies. However, findings from the remaining third that relied on self-reports only may warrant caution as Lerner, Calhoun, Mikami, and De Los Reyes (2012) argued self-reports by students with ASD may be unreliable.

Independent Variables

All interventions described in this review were focused on helping participants develop skills relevant to the demands of university life (e.g., CBT, social skills training), though interestingly, only ten studies described an in-situ component such as a mentor attending a social activity with their mentee. This trend away from traditional academic supports (e.g., scribes, class notes, recorded lectures) reflects a growing awareness of the unique needs of students with ASD (Kuder and Accardo 2017). More research may be required to determine whether in-situ supports are needed, however, as students with ASD have a poor record for generalization (Ncube et al. 2018).

The most compelling issues for many students with ASD are non-academic (Gelbar et al. 2015). Thus, it was not surprising that most of the interventions in this review were focused towards non-academic issues and only two studies had a specific academic focus (Jackson et al. 2018; Ness 2013). While the recognition of the non-academic needs of students with ASD is important, the significance of their academic difficulties and need for academic supports may have been undervalued. Academic difficulties have been identified as causing stress and anxiety Shmulsky and Gobbo (2013) and contributing to the high non-completion rates identified by Newman et al. (2011). Also, Nasamran et al. (2017) found that academic achievement (of high school students) was a predictive factor for post-secondary course completion. In addition, many students with ASD have indicated that they prefer to use academic supports and that academic supports are their most helpful support (Anderson et al. 2018). Indeed, Accardo et al. (2018) found in a survey of 23 respondents that 91% stated that academic coaching was a preferred support, compared with only 30% who preferred peer mentors, 26% who preferred a support group or social skills group, and 9% who preferred self-advocacy training. Thus, while non-academic support research is important, the academic needs and support preferences of many students with ASD suggest there may be a case for more research on academic issues and supports.

Peer mentors and Student Assistants and Therapists

Individualized assistance allows supports to be more flexible and thus can potentially meet a more diverse range of needs (Roberts and Birmingham 2017). Indeed, the peer mentors, student assistants and student therapists in the current reviewed studies assisted mentees with social skills, anxiety, academic achievement, self-efficacy and future orientation (Table 4), and most participants revealed they were satisfied with their programs (Table 1). This conclusion contrasts with the more variable findings of prior research. For example, one participant in Knott and Taylor (2014) described mentoring as humiliating while another stated that their mentor was their most important support. One participant in Simmeborn Fleisher (2012) indicated that working with a stranger was stressful. Further, Accardo et al. (2018) reported that 39% (n = 9) of participants (from their survey of 23 US college students with ASD) had not used or did not plan to use a mentor. The distinguishing feature of the current studies may have been the provision of extensive training and supervision of mentors (a feature not specified in the prior studies), and Ness (2013) suggested that the compatibility of mentors and mentees may also be important. Thus, the current findings suggest the quality of the mentor program may be critical and further research may be required to confirm these suggestions.

Strong technological skills have been identified as a strength of many students with ASD (Anderson et al. 2018). Less than a third of the reviewed interventions, however, incorporated the use of technology even though most participants indicated that they found interventions that used technology enjoyable and useful.

Both group and individual formats were used to implement the interventions. In Furuhashi (2017), Pugliese and White (2014), and Siew et al. (2017) the participants indicated that they preferred group meetings because they provided opportunities for socialization and to share ideas with like-minded peer students who were experiencing similar issues. In contrast, the participants in Ames et al. (2016) rated individual instruction as most helpful because the participants felt more comfortable discussing concerns in a private setting. These differing preferences may reflect the diversity of students with ASD, and the different purposes of the interventions, but they may also highlight the merits of including both formats due to the diversity of needs and student characteristics.

Program Effectiveness

There were insufficient experimental or quasi-experimental studies to classify any intervention as an evidenced-based practice (Reichow et al. 2008), nevertheless, there were some noteworthy findings. Improvements were reported by some participants in all seven experimental and quasi experimental studies reviewed, demonstrating that these supports may have promise for some students. There was much variability of response to intervention between participants (within each intervention), but only Jackson et al. (2018), who examined the use of graphic organizers to teach writing skills, and Koegel et al. (2013) who studied structured social planning in a socialization program, found unequivocal improvement for all participants (Table 2). The heterogeneity of students with ASD, including differing severity of ASD and the presence of comorbid conditions may explain some of this variation (Jansen et al. 2017; Pugliese and White 2014) but, unfortunately, few studies provided such details.

Some researchers may have used insufficient sessions for an effect to be observed. Optimum intervention periods for students with ASD have not yet been identified for any of the interventions reviewed, but it was interesting that of two similar social skills training interventions the longer one (Koegel et al. 2013, 22 weeks) had better outcomes for the participants (Ashbaugh et al. 2017, 10 weeks).

Most interventions were found to be feasible in that they were relatively easy and cheap to implement, but satisfaction with the intervention, which was typically found favorable when reported, was only measured in just over half the studies (Table 2). In some studies, the intervention was rated as helpful in contrast to the objective evidence that showed no or minimal improvement in the dependent variables (e.g., Pugliese and White 2014; White et al. 2016). The recently reported study by Ncube et al. (2018) also reported a divergence between student satisfaction ratings with a peer mentoring intervention and standardized outcome measures of social support or quality of friendships. This may add weight to Lerner et al. (2012) conclusion that self-reports of high functioning (adolescent) students with ASD may be unreliable due to poor ability to recognize or report on their own difficulties. It also highlights the importance of supplementing self-reports and satisfaction ratings with objective empirical findings.

White et al. (2016) pointed out in their study that compared a virtual reality (VR) avatar social skills intervention with a psychosocial cognitive behavioral intervention, that VR was no less effective than more traditional supports, but that VR required fewer trained staff, was more readily accessible, and was potentially more economical. Thus, while the present evidence does not demonstrate that technological interventions have an outcome advantage, technologically based interventions may be considered as a feasible alternative.

As previously noted, pre-experimental studies provided information on the feasibility of a range of interventions and many gave preliminary indication of potential direction for more rigorous research. Specifically, the peer mentoring (e.g., Siew et al. 2017), occupational therapy (Schindler and Cajiga 2015), and transition programs (e.g., McLeod and Harrison 2013; Rando et al. 2016) were generally rated as enjoyable and helpful by their participants who typically also reported improved socialization and strong academic achievement.

Suggestions for Future Research

Most of the reviewed research was pre-experimental and typically small scale. Also, few participant details were provided limiting our conclusions regarding the effectiveness of the interventions for individuals with specific characteristics. Thus, larger scale experimental research with longer intervention periods and better descriptions of participant characteristics (ASD severity ratings and comorbid conditions) is indicated. Collaboration among institutions could be considered to achieve this outcome. Also, replication of some of the more promising reported interventions with experimental or quasi-experimental designs (with larger sample sizes and longer intervention periods) may assist in moving toward best practice guidelines. Given the variability in responses to intervention, rigorous single case experimental designs (which are particularly useful during the initial stages of evaluating promising interventions) may have an important role to play in examining factors associated with the idiosyncratic response to intervention.

There were no reported interventions for sensory sensitivities in the reviewed studies though many students with ASD have indicated this is a concern (Anderson et al. 2018; Knott and Taylor 2014) and it has been recognized in the Diagnostic and Statistical Manual of Mental Disorders, 5th ed. (DSM-5; APA 2013). Sarrett (2018) also found that sensory friendly spaces and practices were the second most desired accommodation that the participants would have liked to have received in higher education and that lack of support for sensory sensitivities diminished the effectiveness of other supports. Thus, there is a need for research on this issue, ideally in consultation with students with ASD. Given Baldwin et al. (2014) found that 46% of adults with high functioning autism and Asperger’s were overeducated for their job classifications, research on employment outcomes following graduation may also be warranted.

Limitations of the Review

The scope of this review was limited to the examination of empirical evidence. Nevertheless, qualitative findings may have been helpful to better understand the variable responses, such as reasons why some participants rated an intervention as helpful when there was no objective evidence of behavioral change. In addition, the review was limited to high functioning students with ASD. Nearly a third of students with ASD have a comorbid intellectual disability, and this review adds no insights into their experiences in post-school education. Further, studies were accepted where some participants did not have a formal ASD diagnosis, though a self-diagnosis may not be reliable.

There were 13 studies (12 from the US and 1 from the UK) that provided details of race/ethnicity and only one that provided details of socio-economic information. These studies revealed an apparent overrepresentation of Caucasian participants (87%) which may limit the generality of the findings and indicate a need to determine why non-Caucasians are not selected as participants. Finally, although the inclusion criteria for our review allowed for unpublished studies, our search found only peer reviewed journal articles and dissertations, so there may have been a publication bias (Schlosser et al. 2007).

Summary and Conclusion

There were 24 studies reviewed which substantially extended the only prior review, but there were insufficient experimental studies for any intervention to be classified as evidence based. Many studies, however, (including the 17 pre-experimental studies that were uninterpretable with respect to causation), demonstrated feasibility and social validity for a range of interventions, including mentoring, transition programs, CBT, academic skills, and social skill development, suggesting they may have promise and warrant further investigation. Participant responses to the interventions were very diverse but few participant characteristics were provided, limiting the analysis of the variable outcomes. Thus, larger experimental studies with better description of participant characteristics are needed to improve the interpretation of intervention results and to increase the external validity of the findings. A lack of research on academic supports was also identified, despite prior research indicating that academic supports are often a preferred service of students with ASD. More research may be needed to quantify the contribution of academic challenges towards the low completion rates of post-secondary students with ASD, and to clarify the importance of academic supports. Researchers may also need to consider the wants, as well as the needs, of students with ASD when choosing and designing intervention research projects.