PC1263: Artificial Intelligence as the Pediatrician's Co-pilot to Strengthen Telehealth Care in Underserved Communities

Purpose/Objectives: The purpose of this study was to assess the accuracy of Artificial Intelligence (AI) technologies as a potential supplement for pediatric telehealth care in underserved communities.

Design/Methods: Knowledge integration was performed across disciplines relevant to pediatric telehealth care in the context of advances in AI-assisted healthcare delivery. Meta-analyses were conducted to determine the accuracy of AI telemedicine in the interpretation of clinical data for the period of 2012 to 2022. This included a review of literature on advances in AI-assisted detection of pediatric pathologies relevant to telehealth care in underserved communities. Using a ‘problem-spaced approach,’ the ethical and technical challenges of implementing AI into remote-presence technologies were assessed.

Results: AI interpretation of patient histories diagnosed pediatric presentations including bronchiolitis, acute asthma exacerbations, pneumonia, roseola, gastroenteritis, sepsis, and meningitis with accuracies ranging from of 82% to 93%. The accuracy of AI-assisted interpretation of visual and auditory data varies across disciplines and organ systems. Through digital images, AI systems identified common pediatric skin lesions with sensitivity of 86% and specificity of 86% and detected over 200 genetic syndromes with sensitivity of 78% and specificity of 93%. Analyzing videos of neonatal hip ultrasounds, AI systems detected developmental hip dysplasia with sensitivity of 84% and specificity of 90%. Using recordings of auscultations, AI systems identified common pediatric murmurs with sensitivity of 91% and specificity of 91%. Overall, diagnoses provided by AI systems were shown to be more accurate than those of junior staff pediatricians but less accurate than senior staff pediatricians. Thematic analyses indicated that ethical challenges of AI-supported telehealth care include transparency, accountability, and informed consent. Technical challenges include digital literacy, data security, and ambiguities in regulatory frameworks.

Conclusion/Discussion: AI has demonstrated the capacity to serve as a valuable asset in detecting common pathologies relevant to pediatric telehealth care. Incorporating AI as a co-pilot for pediatricians has the potential to optimize the quality of pediatric telehealth care by facilitating triage services, promoting early intervention, and increasing the efficiency of virtual healthcare. Implementing AI-assisted virtual care is an important next step toward achieving health equity in underserved communities.