The broader impact / commercial potential of this Small Business Technology Transfer (STTR) Phase I project is to build a condition assessment and monitoring platform that leverages artificial intelligence (AI) for intelligent decision support to expedite diagnosis and facilitate personalized therapy solutions. Currently, patient history data is obtained verbally, and poor health communication between clinician and patient potentially causes 78% of misdiagnoses, resulting in 80,000 avoidable hospital deaths and $750 billion in costs to the US economy each year. Language barriers exacerbate this situation significantly, increasing the risk of severe and frequent adverse outcomes by 49%. Patients with cultural, language, and low health literacy are twice as affected. The proposed work has three major technical objectives: (1) develop an interactive visually-supported healthcare data capture user interface (UI) to overcome communication and language barriers, (2) develop an AI-based risk assessment engine to communicate symptoms to the provider for triage decisions, and (3) leverage predictive analytics to recommend appropriate treatment options and develop a visualization-based condition monitoring and management tool. The proposed data collection and AI-powered provider decision support platform will empower patients and providers by enhancing communication, improving health outcomes, and reducing healthcare costs.This Small Business Technology Transfer (STTR) Phase I project is designed for accessibility across all levels of literacy and is language-agnostic. The platformâs digitally-native, cloud-based approach allows it to be scalable for widespread impact and faster impact. The frontend app captures important patient information and symptoms before the patient exam and summarizes the patient self-report in an easily readable format for the clinician. The clinician will potentially be more prepared, facilitating the delivery of personalized care and improved outcomes by avoiding misdiagnosis, errors, and complications earlier in a patientâs journey. The AI-based model pulls structured data, such as demographic information and symptoms from the patient self-report, and the dataset undergoes supervised learning using historical, real-world outcomes data. Multiple AI models will be evaluated, including classification and neural network models, and those with sufficiently accurate performance will be considered as inputs for ensemble learning, the final prediction output for this project. Example outputs include patient evaluators such as consistencies within the patientâs self-report, an estimated risk assessment score, and an approximate diagnosis accuracy score.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review crit
