Reducing Bias in Predictive Model for Hospital Readmission
S. A. Ponmani, Suryaa Ks, Syed Javith R · 2025
Readmissions in the hospital are one of the most difficult tasks for the healthcare industry. It deals with various constraints such as increased costs, resource deficiency and also affects patients’ mental health. The current systems often fail to address biases present in the training data particularly related to the demographic attributes. This project proposes a bias aware solution aimed at mitigating such disparities while enhancing readmission prediction. The primary objective is to develop an equitable data driven healthcare support tool that ensures fair and efficient patient care. This system is provided with a web interface that allows patients to view their real time health status. It is built with a readmission prediction engine that allows admins to predict the current readmission status of patients. Not only the system predicts but also if the patient is likely to be readmitted, the system will schedule the appointment with the same doctor, extend the patient's bed stay ensuring additional care to them. This system is also embedded with Gemini large language model to provide AI generated explanation. Therefore, the system performs the resource allocation to all the patients in an efficient manner. This project aims to improve patient outcomes, optimize hospital efficiency, and support legal and unbiased healthcare delivery.