Research Symposium

BIO

My name is Harjith Pradeep and I am from Parkland, Florida. I am currently a sophomore at Florida State University majoring in behavioral neuroscience. My current plans for the future are to go to medical school and become a doctor. My current research interests have anything to do with medicine in hopes of advancing current medical research.

Annotation and Information Extraction of Social Determinants of Health from Social Worker Notes of Pediatric Transplantation

Abstract

Electronic health records (EHRs) are widely used to develop machine learning models to predict different health outcomes. However, these models can exhibit inherent biases because they can lack crucial risk factors which are related to the social determinants of health (SDoH), which can contribute up to 40% of preventable deaths. In this project, we annotate SDoH information in de-identified social worker notes in the EHRs of pediatric transplantations at UF Health Shands Children’s Hospital. The first phase of the project is creating a precise annotation framework to ensure there is high quality data being trained by the model. In order to do this, we refine the annotation guideline using a subset of notes so the annotation can be done accurately among all the annotators. The second phase of the project is to benchmark various transformer-based models to identify the SDoH information from all the notes. After the model is fully trained and evaluated, we will also apply generative AI models to extract SDoH information using a few-shot approach. The project will allow healthcare providers to predict health outcomes and identify SDoH factors that are indicative of poor outcomes and support decision making for patients.

Keywords: Social Determinants of Health Annotations