Online Evaluation of the Diagnostic Accuracy of BlinkLab's Digital Assessments for Autism

Summary

This observational study aims to evaluate how patterns of behavioral and sensorimotor responses measured using the BlinkLab Dx1 smartphone application relate to autism diagnoses in children ages 2 to 11. BlinkLab Dx1 is a non-invasive, smartphone-based application under development as a diagnostic aid for healthcare providers assessing autism. In this study, children who have undergone a neurodevelopmental assessment within the past 12 months will complete two short, video-based sessions using the BlinkLab Dx1 app. The app presents visual and auditory stimuli and records reflexive sensorimotor responses and patterns of repetitive behavior. Additionally, primary caregivers will answer a short questionnaire in the app about symptoms and development. Information about prior neurodevelopmental assessments, including documented DSM-5-based diagnoses from routine clinical practice, will be collected retrospectively. The study will examine how the app's neurobehavioral measurements relate to previously assigned clinical diagnoses. These paired data will be used to develop and evaluate a machine learning-based algorithm using separate training and testing datasets to assess whether patterns measured by BlinkLab Dx1 can help distinguish children with autism from children without an autism diagnosis. This study does not involve any treatment or medical intervention.

Detailed Description

Specification of the Time Perspective section in the Study Design: this study has a hybrid time perspective, as BlinkLab Dx1 measures are collected prospectively during remote sessions, while the clinical reference standard (presence of in a DSM-5-based diagnostic report from a neurodevelopmental assessment within the prior 12 months) is collected retrospectively. The clinical reference standard is based on neurodevelopmental assessments conducted in routine clinical care and is collected without influence from study procedures. The study will use paired BlinkLab Dx1 measurements and clinical reference standard diagnoses to develop and evaluate a machine learning-based classification algorithm. The dataset will be divided into separate training and testing subsets, with the training dataset used to develop the model and determine classification thresholds, and the independent testing dataset used to evaluate diagnostic performance. The model will generate a classification categorizing participants as "Positive for autism", "Intermediate" or "Negative for autims". Participants with intermediate results are considered to have indeterminate findings and are excluded from primary diagnostic performance analyses, which are based on participants with definitive positive or negative test results.

Eligibility Criteria

Interventions