The American Academy of Pediatrics has mandated identification of autism (ASD) beginning at the routine 18-month visit, based on evidence that early intervention for children with autism improves outcomes. However, the US Preventive Task Force has not endorsed that recommendation, in part because of limited validity data. This study builds on a series of studies by our group aimed at improving early autism identification. We have developed a promising novel and efficient screener for primary care called Toddler Autism and Development Adaptive Screen (TADAS). The TADAS was developed using machine learning to automatically determine which items to collect from the parent in each case based on data from a community sample in a previous project. Currently used autism screeners have high rates of missing children who have signs of ASD or Developmental Delay (DD) and also problematic and costly rates of over-referral. Data from our prior studies showed that TADAS had nearly three times the sensitivity (.94 vs .32) and similar specificity (.87 vs .90) to the currently recommended and most popular screen for autism (M-CHAT-R/F) and more than twice the sensitivity (.94 vs .39) and similar specificity (.79 vs .85) for developmental delay of the current standard Age & Stages Questionnaire-3tm. The TADAS also provides greater efficiencies since a difficult-to-implement recommended follow-up interview is not required and there is less parent burden in overall number of items. This proposal is to program TADAS for online delivery then validate it against diagnostic testing for ASD and DD in a new community sample of children coming for their 18-month well child visit that better reflects the U.S. population in racial, ethnic and socioeconomic status than our original study population. If our initial results are further validated in this study, screening for early ASD and DD could be much more effective with corresponding improvements in child outcomes through earlier intervention. A side benefit of the project will be an efficient Computer Adaptive Testing language screen based on existing national language abilities data that could be used independently to both track language impaired toddlers who are at risk for later development of autism as well as for general language screening in primary care. These innovative new tools are being created within an online clinical support system, called CHADIS that has additional features to assist in the entire process of early identification, referral, and tracking. CHADIS provides individualized patient education for the parent; clinician use of care coordination software connecting primary care with early intervention services with options to share data and track outcomes; as well as options for doctors to receive required Board re-certification credits for participating in autism screening quality improvement efforts. The final product and "system of care" of value to clinicians, insurers, municipalities and states.
Public Health Relevance Statement: Project Narrative The American Academy of Pediatrics has mandated identification of autism beginning at the routine 18-month visit, based on evidence that early intervention for children with autism improves outcomes, but there are concerns about the accuracy of currently available screening tools at that age. This project builds on data from a previous grant showing that using machine learning algorithms individualizing presentation of items to screen for autism and developmental delay (called TADAS) dramatically improved accuracy of screening with lesser parent burden and benefits to office workflow. The proposed study is required to bring this promising tool to clinical use by programming it into our online clinical process support system (CHADIS) and then completing validation in a new community sample of toddlers weighted to be representative of the U.S. population.
Project Terms: Academy; Age; ages; Algorithms; Certification; Child; 0-11 years old; Child Youth; Children (0-21); youngster; Chiroptera; Bats; Communities; Computers; Diagnosis; Gold; Grant; Insurance Carriers; Insurers; Interview; Equipment and supply inventories; Inventory; Language; Language Tests; Parents; Patient Education; Patient Instruction; Patient Training; Patients; Pediatrics; Physicians; Primary Health Care; Primary Care; Primary Healthcare; Psychometrics; Questionnaires; Recommendation; Research; Risk; Computer software; Software; Specialty Boards; Specificity; Testing; Time; Vocabulary; Vocabulary Words; Work; S Phase; S Period; Synthesis Period; Synthesis Phase; Developmental Delay Disorders; Developmental Delay; Specific Child Development Disorders; Diagnostic tests; Socioeconomic Status; Socio-economic status; socio-economic position; socioeconomic position; Advisory Committees; Task Forces; advisory team; autistic children; children with ASD; children with autism; children with autism spectrum disorder; base; improved; Clinical; Phase; Series; Screening procedure; screening tools; Childhood; pediatric; Early Intervention; Collaborations; tool; machine learned; Machine Learning; Ch'i; Qi; programs; Side; Visit; American; Performance; Municipalities; novel; Toddler; Social Support System; Support System; Early identification; Sampling; response; theories; developmental disorder; developmental disease; Preventive; Autism; Autistic Disorder; Early Infantile Autism; Infantile Autism; Kanner's Syndrome; autistic spectrum disorder; autism spectrum disorder; Consent; Data; Validation; Characteristics; Process; follow-up; Active Follow-up; active followup; follow up; followed up; followup; Development; developmental; cost; care systems; care services; Outcome; Population; innovation; innovate; innovative; case-based; racial and ethnic; ethnoracial; demographics; service intervention; evidence base; data sharing; screening; learning strategy; learning activity; learning method; improved outcome; language impairment; Preventive service; Preventative service; study population; recruit; early screening; Medicaid services; care coordination; coordinating care; machine learning algorithm; machine learned algorithm; data repository; Data Banks; Databanks; data depository; online delivery; delivered on-line; delivered online; on-line delivery; Well Child Visits
