Developing AI-Derived Multilevel Risk Scores for Oral Cavity and Oropharyngeal Cancer Patients - Modified Project Summary/Abstract Section ABSTRACT Each year, oral cavity and oropharyngeal cancers (OCC/OPC) claim the lives of over 12,000 Americans. Despite modest improvements in five-year survival rates, significant gaps remain in timely diagnosis and access to high-quality treatment. These gaps are especially prominent among individuals facing higher levels of socioeconomic risk, particularly black individuals, people with low incomes, as well as those residing in rural areas. Existing research has not fully characterized how the combined effects of individual and neighborhood-level social factors contribute to treatment and survival outcomes. Our central hypothesis is that multilevel social risk is a key driver of delayed treatment and increased mortality in OCC/OPC. This proposal will comprehensively evaluate upstream factors (economic stability, education access, healthcare access and quality, neighborhood-built environment, and social-community context) and use Artificial Intelligence (AI) and Machine Learning (ML) tools to develop and validate outcome-specific social risk scores to predict treatment and outcomes. We will use real-world electronic health record data from ~4,000 adult patients (aged 18 and older) diagnosed with OCC/OPC in the University of Florida Health System. These real-world data (RWD) will be linked with several external datasets (e.g., American Community Survey, Area Health Resources File) to evaluate outcomes, including treatment delay, receipt of guideline-concordant care, and mortality. We propose a study among patients diagnosed with primary OCC/OPC between 2012–2024. In Aim 1, we will build an RWD cohort of OCC/OPC and link multilevel datasets. In Aim 2, we will develop ML- based social risk management algorithms to predict OCC/OPC treatment and mortality. Findings will provide the foundation for scalable, patient-centered interventions that address social risk. Our future R01 will externally validate these risk scores, apply causal inference methods, collect patient-reported outcomes, and engage stakeholders to co-develop strategies for improving OCC/OPC care delivery and health outcomes across at-risk populations.
