Integrating Large Language Models (LLMs) and Vector Databases into Healthcare Operations

The medical field, especially in US, has transitioned from paper-based records to Electronic Health Records (EHRs). Switching to EHRs have offered health care providers several advantages over paper-based records, such as coordination, documentation, security, and communication with patients. But as EHRs were primarily designed for billing and documentation, they have led physicians to do more administrative tasks, instead of healthcare related tasks. Despite excellent adoption of EHRs, there have been very few studies addressing the side-effects related to EHR usage, such as - Physician burnout, poor user interface, complex usability, and time incurred to perform clerical tasks. In this study, we propose a prototype where the patients can interact with the LLM enabled system integrated with vector database. This system will allow patients to get initial observations and physician-like results to their queries that don't need immediate escalation to physicians. The experiments and results section demonstrates how the proposed study can help in reducing the clerical burden and cognitive overload of clinicians.