Neural Engineering Techniques for Autism Spectrum Disorder, Volume 2: Diagnosis and Clinical Analysis

Neural Engineering for Autism Spectrum Disorder, Volume Two: Diagnosis and Clinical Analysis presents the latest advances in neural engineering and biomedical engineering as applied to the clinical diagnosis and treatment of Autism Spectrum Disorder (ASD). Advances in the role of neuroimaging, magnetic resonance spectroscopy, MRI, fMRI, DTI, video analysis of sensory-motor and social behaviors, and suitable data analytics useful for clinical diagnosis and research applications for Autism Spectrum Disorder are covered, including relevant case studies. The application of brain signal evaluation, EEG analytics, fuzzy model and temporal fractal analysis of rest state BOLD signals and brain signals are also presented.

A clinical guide for general practitioners is provided along with a variety of assessment techniques such as magnetic resonance spectroscopy. The book is presented in two volumes, including Volume One: Imaging and Signal Analysis Techniques comprised of two Parts: Autism and Medical Imaging, and Autism and Signal Analysis. Volume Two: Diagnosis and Treatment includes Autism and Clinical Analysis: Diagnosis, and Autism and Clinical Analysis: Treatment.

• Presents applications of Neural Engineering techniques for diagnosis of Autism Spectrum Disorder (ASD)
• Includes in-depth technical coverage of assessment techniques, such as the functional and structural networks underlying visuospatial vs. linguistic reasoning in autism
• Covers treatment techniques for Autism Spectrum Disorder (ASD), including social skills intervention, behavioral treatment, evidence-based treatments, and technical tools such as Magnetic Resonance Spectroscopy for ASD
• Written by engineers for engineers, computer scientists, researchers and clinicians who need to understand the technology and applications of Neural Engineering for the detection and diagnosis of Autism Spectrum Disorder (ASD)

• Language: English
• Publisher: Academic Press
• Release date: Oct 17, 2022
• ISBN: 9780128244227

Book preview

Part 1

Autism and clinical analysis: Diagnosis

Outline

Chapter 1 Remote telehealth assessments for autism spectrum disorder

Chapter 2 Maternal immune dysregulation and autism spectrum disorder

Chapter 3 Reading differences in eye-tracking data as a marker of high-functioning autism in adults and comparison to results from web-related tasks

Chapter 4 Parents of children with autism spectrum disorders: interventions with and for them

Chapter 5 Applications of machine learning methods to assist the diagnosis of autism spectrum disorder

Chapter 6 Potential approaches and recent advances in biomarker discovery in autism spectrum disorders

Chapter 7 Detection and identification of warning signs of autism spectrum disorder: instruments and strategies for its application

Chapter 8 Machine learning in autism spectrum disorder diagnosis and treatment: techniques and applications

Chapter 9 Inhibition of lysine-specific demethylase 1 enzyme activity by TAK-418 as a novel therapy for autism

Chapter 10 Behavioral phenotype features of autism

Chapter 11 Development of an animated infographic about autistic spectrum disorder

Chapter 12 Fundamentals of machine-learning modeling for behavioral screening and diagnosis of autism spectrum disorder

Chapter 13 A comprehensive study on atlas-based classification of autism spectrum disorder using functional connectivity features from resting-state functional magnetic resonance imaging

Chapter 14 Event-related potentials and gamma oscillations in EEG as functional diagnostic biomarkers and outcomes in autism spectrum disorder treatment research

Chapter 1

Remote telehealth assessments for autism spectrum disorder

Angela V. Dahiya¹, Jennifer R. Bertollo¹, Christina G. McDonnell² and Angela Scarpa¹,    ¹Department of Psychology, Virginia Tech, Blacksburg, VA, United States,    ²Department of Psychology, University of Wyoming, Laramie, WY, United States

Abstract

Although early identification of autism spectrum disorder is critical for ensuring timely diagnosis and access to support, striking disparities and inequities delay the diagnostic process. Technology approaches hold exciting potential for reducing barriers to care and increasing access to autism diagnosis. In the current chapter, we review evidence for the use of telehealth-based technology for facilitating autism diagnostic assessments, including (1) live videoconferencing, (2) asynchronous video observations (i.e., nonlive observations of current behavior via video), (3) retrospective video analysis (i.e., nonlive observation of past behavior), (4) mobile and web applications, and (5) online websites and phone interviews. We conclude with limitations of this research and directions for future research in this critical area of inquiry.

Keywords

Autism spectrum disorder; assessment; telehealth; technology-based assessment

1.1 Introduction

Autism spectrum disorder (ASD), diagnosed in about 1 in every 54 children in the United States, is a neurodevelopmental disorder characterized by two domains of diagnostic criteria described in the Diagnostic and Statistical Manual of Mental Disorders-5th Edition (DSM-5; [1]): persistent difficulties with social communication and reciprocal interactions, and the presence of restricted and repetitive behaviors or interests (RRBs). Social communication difficulties may include difficulties with socio-emotional reciprocity (e.g., reduced or different ways of initiating or responding to social interactions and sharing interests or emotions with others; reduced or different conversational styles), nonverbal communication behaviors (e.g., differences in the use of eye contact, gestures, facial expressions, etc.), and difficulty with relationships with others [1]. RRBs may include stereotyped or repetitive behaviors (e.g., motor movements, object use, or speech), insistence on sameness (e.g., significant difficulty with transitions, strong preference for routines), restricted or fixated interests, or differences in sensory responses and interests [1]. These core domains can impact various aspects of day-to-day functioning, including social communication, relationships, employment, adaptive behavior skills, and quality of life.

It is well documented that early supports lead to greater independence and quality of life for autistic¹ individuals [2], but access to services is hampered by delayed identification and diagnosis [3]. In fact, research indicates that ASD can be reliably diagnosed as early as two years of age, but the median age of first diagnosis in the United States is 4 years and 3 months [4]. In addition to racial, ethnic, and sex and gender disparities in the timeliness of autism identification (see [5], for review), children in remote or rural communities, and those below the poverty line, are diagnosed significantly later than those in urban or more affluent communities [6,7]. Barriers such as geographic isolation, financial instability, lack of local resources, and more recently, the COVID-19 pandemic that requires physical distancing, are challenges that many remote communities continue to face [8,9]. Even poor urban communities face a similar lack of ASD provider availability compared to wealthier communities [10], and all individuals may face uncontrollable circumstances that create access barriers regardless of locale (e.g., natural disasters, snowstorms, illness). As such, remote assessment opportunities for ASD diagnosis widen our capacity to reach as many people as possible when they are unable to come to a clinic in person or simply do not have access to experts in their community.

1.1.1 In-person standardized assessments for autism spectrum disorder

Prior to formal diagnostic assessment for ASD, screening methods are intended to catch characteristics of ASD early in development and subsequently refer a child for a thorough diagnostic assessment if autistic features are present. The American Academy of Pediatrics recommends universal autism screening for all children throughout infancy and toddlerhood; specifically, they recommend all children be screened for broad behavioral and developmental concerns at 9, 18, and 30 months, and specifically for ASD using a standardized screening tool at 18- and 24-month well visits [11]. However, adoption of these recommendations has been inconsistent, as not all governing organizations put forward the same screening recommendations and resulting policy mandates [12]. One common route of universal screening is to have primary care physicians or staff administer an evidence-based screening measure such as the Modified Checklist for Autism in Toddlers–Revised with Follow-up (discussed in more detail later in this chapter) during child well visits, in order to increase the chances of flagging children with developmental, social, or communication concerns who may otherwise go unnoticed until beginning preschool or Kindergarten. Once a screening instrument or qualified professional (e.g., pediatrician) identifies characteristics suggesting increased likelihood of being on the autism spectrum, the child is then referred for a comprehensive diagnostic assessment.

Currently, standard face-to-face ASD diagnostic assessments consist of several hours of testing, including a developmental history interview with one or more caregivers (e.g., Autism Diagnostic Interview-Revised, ADI-R; [13]) and an observational behavioral assessment with the individual suspected to meet criteria for a diagnosis of ASD (e.g., Autism Diagnostic Observation Schedule, 2nd Edition, ADOS-2; [14]). At present, many experts consider these two measures to comprise the gold standard protocol for an ASD evaluation and thus they are both widely used instruments. The ADI-R gathers necessary medical and developmental information, while evaluating social communication (e.g., stereotyped utterances, little use of nonverbal communication), reciprocal interaction and peer relationships (e.g., limited response to others, lack of reciprocal conversation abilities), and RRBs (e.g., presence of preoccupations, complex body mannerisms, sensory interests). The ADOS-2 uses specific interaction tasks to prompt for the aforementioned social communication differences and RRBs, which a trained clinician facilitates, observes, and codes. In addition to the ADI-R and the ADOS-2, an assessment battery often includes measures of cognitive and language abilities to further specify any co-occurring intellectual or language impairment. Multidisciplinary evaluations also incorporate speech-language assessments, school-based reports, or medical consultations [15]. Finally, parent- or caregiver-report of adaptive behavior functioning (i.e., whether an individual has developed the age-expected skills necessary to function in daily life) and other characteristics (e.g., restricted interests, sensory differences, emotion dysregulation, anxiety) are also typically collected to strengthen diagnostic decisions and recommendations for services or accommodations.

Professionals recommend administration of this or a similarly comprehensive protocol in order to screen and diagnose ASD, but many providers may only utilize one of the methods due to the length of time, cost, and required training to administer an involved battery of interview and observational methods [16]. Such a diagnostic protocol requires several hours of direct face-to-face contact, as well as a significant amount of resources in order to train providers on the correct administration and scoring techniques of these assessments. Thus, people seeking a diagnosis of ASD are often on long waitlists or left undiagnosed [17]. With this context in mind, it is imperative to explore efficient tools that can decrease time and cost, while increasing accessibility and implementation of screening and diagnostic assessments.

Several recent efforts have been made to overcome time and resource barriers by developing shorter observation measures that can be administered by community providers without substantial training (i.e., [18,19]). However, technology-based or telehealth modalities can provide another viable and more accessible alternative to in-person diagnostic evaluations. Previous research has examined technology as a means to deliver assessment services for autistic children in a cost-effective way [20], which is consistent with current telehealth therapy practices [21,22] and will be further explored throughout this chapter.

1.1.2 Significance of remote assessments for autism spectrum disorder

Currently, the critical need for remote assessments for the screening and diagnosis of ASD is highlighted by the COVID-19 pandemic, which has necessitated social distancing, quarantining, and government-issued stay-at-home orders across the country and world. As a result, many clinics and providers were forced to pause their services until they were either permitted to see clients face-to-face, or were able to develop and implement alternative means of reaching families remotely. In the United States, in-person operations were halted altogether for several months and many providers still have not returned to the fully in-person operations they relied on prior to COVID-19, instead maintaining full to partial telehealth-based service options.

Even prior to the pandemic, the average waitlist to be seen by an autism-specific provider spanned from several months to more than a year for a comprehensive diagnostic evaluation, depending on location [23]. Although research consistently supports that ASD can be diagnosed reliably in children as young as 2 years of age and parental concern may arise even earlier, children in the United States are not diagnosed until after four years of age on average, one-third of autistic children have still not been diagnosed by 8 years of age, and many individuals do not receive a diagnosis until adolescence or even adulthood [4,24]. These facts are particularly alarming given the well-documented benefits of receiving supports or services prior to the age of three, and even as young as 18 months of age, for improving long-term quality of life and adaptive outcomes for autistic individuals [2,25]. Further, because many aspects of service access such as insurance coverage and public school accommodations depend upon a formal diagnosis of ASD, the importance of a timely diagnostic assessment cannot be overstated. During the COVID-19 pandemic, already long waitlists have increased further.

Although the COVID-19 pandemic has resulted in an unprecedented need to utilize remote assessment measures, many other circumstances can create barriers to receiving in-person services for families outside of the context of the current pandemic. For example, individuals or families living in rural areas or who are otherwise geographically isolated from major medical centers may not have the ability to travel to a clinic during normal business hours without requesting a full day or more off from work, a loss of pay that may not be affordable to many. Rural areas also see a stark paucity of resources and providers broadly, but particularly those knowledgeable in ASD [9]. These barriers together result in a later average age of diagnosis in rural areas compared to urban or suburban regions [26], which in turn lead to delays in receiving desired supports and/or services that support independence and quality of life for autistic individuals [6,7]. Additionally, caregivers may avoid scheduling services for themselves or their children if they have other children at home and cannot afford or find childcare, as there may not be room or resources for those children at a typical provider's office. Further, several of these barriers such as low socioeconomic status, unmet childcare needs, and difficulty navigating service systems disproportionately affect racial-ethnic minority families and their children, who are significantly less likely than their white counterparts to have their autistic characteristics documented in a formal diagnosis [27]. In sum, the need to understand and utilize the most accessible methods of service delivery is of the utmost importance well beyond this current time of crisis, and novel methods of assessment delivery are necessary in order to increase equitable service access.

Remote telehealth assessment is one such solution that may help to expedite ASD screening and diagnosis in rural communities, where delays are too often the norm, but also in any under-resourced communities or otherwise hard-to-reach and isolated populations [10]. In an increasingly technologically driven world, it is important to consider how such technologies can be leveraged to improve access to screening and assessment practices. We now review the existing evidence for the use of telehealth-based technology for facilitating autism diagnostic assessments, including (1) live videoconferencing, (2) asynchronous video observations (i.e., nonlive observations of current behavior via video), (3) retrospective video analysis (i.e., nonlive observation of past behavior), (4) mobile and web applications, and (5) online websites.

1.2 Telehealth assessments

Several forms of technology can be used to aid in the diagnostic assessment of ASD. Certain aspects of in-person ASD assessment, such as parent or caregiver interviews, can be translated most easily to remote conduct, as having a conversation by phone or video is nearly identical to having the same conversation in-person. Further, it has been demonstrated that phone-based screening methods for this population can help categorize individuals who may be at an increased likelihood of having developmental delays [28,29], prior to a more comprehensive diagnostic assessment. Additionally, more peripheral aspects of comprehensive ASD assessments, such as cognitive and academic achievement testing via telehealth, have been demonstrated to be a valid means of assessing cognitive function without major shifts in scores during COVID-19 [30], although these measures are still understudied in minority populations to ensure their equitable use [21]. However, other core aspects of the standard in-person assessment of ASD, particularly the observation of children or adults presenting with autistic features, require more careful consideration and study through remote platforms. For example, because ASD is characterized by social and communication differences, individuals may have seemingly improved communication abilities if observed with familiar family or caregivers in the home environment as compared to observation in an unfamiliar clinical setting when the clinician may be a complete stranger. On the other hand, certain stereotyped behaviors and sensory reactivity may be more apparent in a home environment, where some individuals may not make the same effort to inhibit their behaviors as they may in public. Finally, if we consider interacting with another adult or child in-person vs through a computer screen or telephone, several aspects of normal social interactions may be less natural or more difficult to assess. For example, the give-and-take of conversations can be much more difficult to navigate via a video meeting, as it is more difficult to know when the other person is done speaking or is preparing to speak, particularly with lagging internet connection. Further, constructs such as eye contact can be difficult to capture accurately, as it is difficult to know whether an individual is looking at their interaction partner, at the camera, or at something else altogether. Moreover, certain tasks such as interactive play with a child cannot be naturally conducted between two individuals on different computers. Therefore, assessment of children's play and interaction skills will typically require a caregiver or family member to interact directly with a child while a provider watches from afar, again changing the context of the interaction drastically from the usual assessment with an unfamiliar provider in an unfamiliar clinic setting.

Despite these potential challenges to remote observation, much recent work has begun to document the feasibility and validity of utilizing remote measures to screen and assess for ASD in children and adults. Videoconferencing, video observations, and video analysis can investigate a range of features and behaviors relevant to the diagnosis of ASD, including gestures, nonverbal and verbal communication, developmental skills, social orienting, object use, and social play. Mobile and web-based applications and resources allow caregivers to report autistic traits and behaviors from their phones or other devices (e.g., laptop, tablet). One additional advantage of assessing individuals remotely through these modalities is that it allows providers to capture behaviors in a naturalistic setting such as their home or community, which can provide a unique and accurate perspective of an individual's behaviors. Two separate systematic reviews [31,32] identified a total of 16 technology-based tools utilized for either screening or diagnostic evaluations for children suspected to have ASD. These reviews focused exclusively on children and did not include those over the age of 12. The current chapter will summarize these reviews, broaden the findings by expanding to studies in adolescent and adult samples, and will synthesize the findings to summarize the current state of the literature and to make practical recommendations for clinical practice.

1.2.1 Videoconferencing (live/in vivo)

Out of the various types of remote technology, live videoconferencing and video observations may be the most comparable to in-person assessments, as they provide naturalistic in-home observations while interacting with providers and clinicians in real time. These interactions are then transmitted via video for an expert to review. There are several telehealth tools that have been implemented across the lifespan, some of which can be facilitated by caregivers instead of clinicians.

Ciccia and colleagues [65] examined videoconferencing as a screener for neurodevelopmental disabilities by utilizing a method known as the Integrated Valuation of Ecosystem Services and Tradeoffs (INvesT) model. This allowed clinicians to conduct parent interviews and speech and language testing via live video calls for children up to 6 years of age to screen for potential neurodevelopmental disorders, including ASD.

A caregiver-facilitated evaluation tool, the TELE-ASD-PEDS [33], is a tele-based version of the ASD-PEDS [34]. The TELE-ASD-PEDS capitalizes on the use of toys and materials already available in the average family's home to allow clinicians to observe behaviors associated with ASD remotely and takes only 15 to 20 minutes to administer. Importantly, the TELE-ASD-PEDS is limited to use in toddlers and young children, up to 36 months of age. The psychometric properties of this measure are currently being evaluated, but data are not yet published.

Evidence-based tele-assessment measures for ASD for older children, adolescents, and adults are even scarcer. A scoping review published in July 2020 [35] identified only ten studies regarding ASD tele-assessment, which primarily utilized live or retroactive video observation. Of these ten studies, only two focused on adolescence to adulthood (i.e., [36] and [37]). Parmanto et al. [36] specifically created a system that integrates videoconferencing with the presentation of images and videos to assess for ASD; their platform uses recording and electronic scoring to aid in the ease of tele-assessment delivery. Parmanto et al. translated ADOS-2 tasks to their tele-assessment tool where possible and stated that their assessment mimics a fluid natural interaction, which is crucial for accurate assessment of the social and communication skills often lacking in ASD. Schutte et al. [37] assessed the usability and reliability of the ADOS-2 Module 4, the module intended for use with verbally fluent adolescents and adults, remotely via a computer. They utilized this measure in 23 adults (ages 19–30 years) with an ASD diagnosis and compared the results to each person's in-person ADOS-2. While diagnostic classification was highly reliable between formats, the actual scores were less reliable (Kappa >0.61 for 21 out of 31 items). Nonetheless, participants were highly satisfied with their remote ADOS-2, suggesting that this is an acceptable format for adult clients to interface with for clinical assessment services.

Additional videoconferencing measures are still under investigation for their utility in remote assessment, but have begun to be used during the current pandemic. The Childhood Autism Rating Scale, Second Edition, Observation (CARS-2Obs; [19]), for example, is a 15-minute interaction that is intended to be observed in person by a professional to confirm an ASD diagnosis, and it consists of eight coded items. This measure shows high correlations with the ADOS-2 (r=0.68), supporting this observation as a tool to differentiate children with and without ASD. The cut-off score of 16 provides specificity of 95.8% and a sensitivity of 62.3%. In light of the current pandemic, several clinics, including our own, have begun to utilize the CARS-2Obs as a parent- or caregiver-facilitated observation, live coached and observed remotely by clinicians and coded according to the same eight items, immediately after the observation. However, it is important to note that the CARS-2Obs has not yet been validated using remote video observations.

Similar to the CARS-2Obs, the Brief Observation of Symptoms of Autism (BOSA; [38]) is an observation facilitated by caregivers with their child that can be used to as part of a protocol to diagnose ASD. It utilizes a set of activities that have been adapted from the ADOS-2 and are facilitated over the course of 15 minutes. Clinicians are able to observe either remotely through a video platform or live in-person with sufficient social distance (i.e., from greater than six feet away or through a one-way mirror). A clinician then codes the individual's skills and behaviors based on the ADOS-2 scoring protocol, which is then mapped onto the DSM-5 checklist for ASD, along with behaviors observed outside of the BOSA administration. Alternatively, this observation can be recorded and coded at a later time. The BOSA has clear benefits in its brevity, ability to be conducted remotely or in clinics, and use of the same items on the ADOS-2, which many professionals consider to be the current gold standard in ASD observational assessment. However, the prescribed materials required for administration of the BOSA are a limiting factor to flexibility and accessibility of this measure. This is easily overcome if conducted in the clinic setting, but remote observation from a client's home would entail either shipping or delivering the materials to each family and then collecting and disinfecting them for the next family. By comparison, the major benefit of the CARS-2Obs is its flexibility of materials. The CARS-2Obs does not have any set materials and instead capitalizes on the toys and games that families already have in their homes, allowing families to participate regardless of their distance from a clinic or a clinic's resources to get materials to families' homes.

According to best practices for ASD assessment [39,40], both of these measures (i.e., the CARS-2Obs and the BOSA) should ideally be used along with other standardized ASD measures (e.g., the ADI-R or other parent-report measures) as part of a diagnostic battery where results are synthesized to form a diagnostic judgment by a licensed professional. This use of other measures in the assessment process is even more important in the case of these abbreviated observational instruments, especially considering that there are minimal empirical data thus far to support their validity. Continued research on these measures will be necessary to validate their use over a telehealth platform (e.g., Zoom, Skype) or through an observation window.

Video conferencing methods thus far have provided promising results for administration of assessments via telehealth, suggesting that remote tele-assessments may be a viable alternative to in-person contact. Additionally, training caregivers to facilitate assessment observations can minimize the need for in-person contact with clinicians to accommodate current safety restrictions, as well as long-term barriers to traveling to a clinic.

1.2.2 Asynchronous video analysis: current

While live videoconferencing methods are able to provide in-the-moment synchronous observations of an individual's behaviors and social interactions, recorded videos afford clinicians the ability to carefully view and analyze asynchronous videos of naturalistic observation without the possible interference of a clinician being present.

As reviewed in Dahiya et al. [31], several examples of this approach are summarized here. One example, investigated by Smith et al. [41], is the Naturalistic Observation Diagnostic Assessment (NODA). Researchers asked caregivers to prompt their children during social interactions in a way similar to how clinicians administer diagnostic observations (e.g., ADOS-2), such as saying the child's name or interacting with them playfully. The caregiver was instructed to record a total of four 10-minute videos, which were then later evaluated by trained clinicians. Similarly, Tariq et al. [42] used a series of videos of an even shorter length to identify possible early signs that may be characteristic of ASD. A follow-up study [43] used a similar method with children in Bangladesh, in which the researchers attempted to identify ASD traits. Results noted positive test accuracy of ASD diagnosis across both settings in all classifiers on the ADOS-2 (85% or above). However, although the sensitivity was positive for both settings, the specificity differed based on the classifier (United States ≥ 50% for only three classifiers; Bangladesh=77% overall). This discrepancy suggests that there could be a difference in how some of the classifiers are administered. Additionally, the variation in video length could impact the difference between each participant and what the video rater analyzes. In the NODA application, the child's videos were limited to 10 minutes and were tagged using five continuous variables via a specific protocol that the raters were trained to code. In the machine-learning protocol from Tariq et al. [43], upwards of 30 features could be coded, suggesting that outcomes could be highly inconsistent due to both the limited length of a submitted video and the number of features that a rater is aiming to analyze during that short period of time.

Chambers et al. [44] also explored the use of video observations and diagnostic coding in a nonEnglish speaking South African population with administration being conducted in their native language of isiZulu. These videos were collected by speech-language pathologists among children at increased likelihood of an ASD diagnosis from 12 to 48 months of age in a natural home environment, in addition to administering other questionnaires including the Early Screening for Autism and Communication Disorder (ESAC; [85]) the Communication and Symbolic Behavior Scales-Developmental Profile Behavior Sample (CSBS; [45]) and the Systematic Observation of Red Flags of ASD [46]. In terms of the implementation of this observation, both the United States and South African teams established 100% agreement among participants, improving the accessibility and adaptability of this video-based method in a different language and/or country.

Although video observations can be used diagnostically to assess for ASD, it is also feasible to observe videos as a screening tool. Considering the importance of adapting these assessment methods to diverse populations, the study conducted by [86] is important to highlight, as this research team applied the video observation method to a sample of families of various socioeconomic status and backgrounds. Children with suspected ASD or language delay (LD) were compared to non-ASD or non-LD groups, and they were recorded during an ADOS-2 administration. Several standard autistic behaviors were specifically coded: social skills, vocal sounds or expressive language, play behaviors, and response to name. Although the sensitivity rate for detecting ASD was relatively low (61%), the specificity of ruling-out ASD was promising (82%). This finding suggests that this method can be effective for screening to differentiate possible ASD-specific from non-ASD concerns [47].

Prior to the COVID-19 pandemic, the Italian Ministry of Health's Early Bird Diagnostic Protocol for Autism Spectrum Disorder project was underway to determine the best screening and diagnostic procedures by age for those with suspected ASD. In response to the COVID-19 pandemic, Conti et al. [48] transitioned the study to enroll toddlers in their remote surveillance protocol (RSP). Although still under study to better understand the utility of this protocol for detection and timely initiation of supports and services, the authors describe their RSP procedures in their recent publication. The RSP begins with a brief parent-child play interaction, which is video recorded and then discussed among the clinical team according to the items on the Toddler Module of the ADOS-2. Parents then participate in three online interviews to assess their child's history and autistic traits, adaptive skills, and social and emotional functioning. After all of this information is collected, the team makes a decision about whether to provide feedback online to the parents about developmental concerns, or whether a live face-to-face visit is necessary to complete the evaluation before providing a diagnosis.

1.2.3 Asynchronous video analysis: retrospective

Another form of technology that has long been used for identifying signs of ASD is analysis of prior video recordings. Frequently, retrospective video analysis examines whether children who are later diagnosed with ASD can be differentiated from other children in their early years of life. Much of the earliest literature on technological ASD identification utilizes family home videos to investigate signs of ASD. The evidence gleaned from retrospective video analysis studies clearly establishes that technology can be useful in identifying signs of ASD, such as joint attention difficulties, limited eye contact, and social hypo-responsiveness.

For example, Baranek [49] examined retrospective videos of autistic children, children with developmental disabilities (DD), and a comparison group of nonautistic children without DD and found that video raters were able to correctly classify children into diagnostic groups 93.75% of the time. Furthermore, in a small pilot study, Adrien et al. [50] found that early signs of ASD such as eye contact and poor variability of emotional expression could be identified in infancy. Osterling et al. [51] showed that ASD could be differentiated from intellectual disability at the first birthday; specifically, autistic children looked at others and oriented to their names less frequently than children with intellectual disability in these home videos.

Another retrospective video analysis study by Clifford and Dissanayake [52] found that autistic children had less joint attention during the second year of life, and the authors suggested that precursors to the joint attention difficulty, such as differences in gaze and affect, may be recognizable as early as 6 months. Similarly, in 15-minute video segments taken between the ages of 12 and 24 months, Clifford et al. [53] showed that several behaviorally coded items differentiated autistic children from their nonautistic peers. These items included peer interest, gaze aversion, anticipatory postures, and proto-declarative showing. Regarding early sensory differences in autistic children, Freuler et al. [54] found that infant home videos of autistic children demonstrate hyporesponsiveness to sensory input, as well as sensory repetitions (such as spinning). Motor differences are also observable in infant home videos; Ozonoff et al. [55] found that gross motor maturity was delayed for children with DD or ASD without regression in these skills compared to nonautistic children and autistic children with regression. Based on this evidence that specific early signs of ASD are identifiable in retrospective video analysis, one study [56] utilized a prospective design to identify specific patterns of gesturing between children who were later more likely to fall above the autism cutoff on the ADOS versus within the autism spectrum or nonspectrum ranges [57].

Retrospective video analyses also predict other outcomes for autistic children, further highlighting the utility of technology in ASD assessment. One study [58] found that the mean level of social communication behaviors in infants later diagnosed with ASD, measured using retrospective videos, predicted Vineland Communication scores and intellectual functioning at ages 3–7 years. Similarly, Receveur et al. [59] found that behaviors coded from video at various time points in early childhood (i.e., 10–12 months, 16–18 months, and 24–26 months) could differentiate children with developmental quotients above and below 50. Overall, these findings largely suggest that signs of ASD can be reliably identified via video recording, which has provided an empirical foundation for tele-assessment and tele-screening practices.

1.2.4 Mobile applications

With over 5 billion people connecting to mobile services during the past year and a projected surge of at least one billion additional users over the next few years [60], mobile applications have become increasingly accessible worldwide. In light of the limitations of the COVID-19 pandemic, mobile applications are progressively being designed to address barriers to care and thus have the potential to bridge the research-to-practice gap in screening and assessment of ASD, especially if facilitated by caregivers.

Duda, Daniels, & Wall [61] evaluated a mobile screening application, the Mobile Autism Risk Assessment (MARA), which utilized multiple choice questions to gather information from caregivers on social skills, communication, and RRBs. This study found that this mobile method had a sensitivity of 89.9% and specificity of 79.7% for future diagnosis of ASD. Similarly, Maleka et al. [62] investigated a mobile application in South Africa, in which the mobile version of the Parents Evaluation of Developmental Status (PEDS) tool had high agreement with a pen-and-paper version when completed by community health workers. Finally, [87] created an online version of the aforementioned model by [63]; the INvesT model in which researchers categorized the likelihood of autism in 12- to 36-month-old children based on caregiver reports of specific developmental concerns. This tool was able to provide data on the likelihood of developmental delays, noting several children who achieved high scores on the measure, which accurately aligned with their diagnosis from an in-person assessment.

An additional app has been piloted in children 18 to 48 months of age, not to assess ASD more broadly, but specifically to measure a child's response to their name, a skill that is often disrupted in young autistic children [88]. In this study, parents would say their child's name, video record their child's response from their smartphone, and also indicate whether they thought their child responded. The purpose of developing this app was to address shortcomings of current ASD assessment measures that either rely on parent/caregiver report or require clinicians to make gross decisions about a child's response to name (i.e., determining whether the child looks in response to their name overall, when only given one or very few trials to observe this behavior), thereby under-appreciating the variability in children's behavioral responses. To this end, Thomas and colleagues found no significant differences in clinician-coded response to name between children with and without diagnoses of ASD or non-ASD developmental delays after one trial. After three trials, autistic children and typically developing children's performance were significantly different. However, 10 trials were necessary before autistic children and those with another developmental delay could be differentiated, and they remained significantly different through 20 trials. While not a diagnostic tool in and of itself, this application is a promising remotely administered and parent-facilitated tool that can aid in ASD screening and in providing more nuanced metrics of social and communication