Hemodialysis is a common treatment for End-Stage Renal Disease (ESRD). To filter blood, it requires vascular access procedures such as thrombectomy and fistuloplasty, which carry elevated risk of infection, cardiovascular events, and mortality. This project aims to develop a model that predicts peri-procedural morbidity, readmission, and mortality for hemodialysis-related interventions, filtered from the ACS-NSQIP surgical database using CPT and ICD codes. Models including Random Forest and TabPFN will be trained and evaluated on AUROC, precision, recall, specificity and sensitivity. Results will be stratified by age, race, and sex to ensure fairness and transparency. The goal is to improve care by helping clinicians identify high-risk patients and deliver personalized, timely care.

I’m participating in SuperUROP to gain experience applying machine learning to real clinical data. My background in AI and biomedical coursework prepared me for this research, and I’m excited to learn how data-driven tools can improve patient outcomes and support clinical decision-making.