Design and Implementation of an AI-Powered Chatbot for Real-Time Health Benefits Navigation with HIPAA Compliance

Abstract

The application of artificial intelligence (AI) in healthcare is on the rise. There is a lack of clinical evidence for the effectiveness of patient-focused chatbots, despite their increasing availability. The rapid advancement of digital healthcare technologies has transformed patient–doctor interactions, yet many individuals still face delays or inaccuracies in obtaining medical guidance due to limited access to healthcare professionals and the reliance on traditional keyword-based chatbots that struggle to understand natural language nuances. In response to this challenge, this work presents the design and implementation of a HIPAA-compliant AI-powered medical chatbot that leverages semantic similarity for real-time query–response retrieval. The methodology involves preprocessing a curated subset of 70,000 patient-doctor interactions from a Kaggle dataset, including text cleaning, stop-word removal, and feature consolidation, followed by encoding patient queries and doctor responses into 384dimensional embeddings using the transformer-based Sentence Transformer model (all-MiniLM-L6-v2). Cosine similarity is then used to retrieve contextually relevant doctor responses efficiently. Experimental results demonstrate that the system achieves an accuracy of 97.4%, recall of 97.4%, precision of 100%, F1-score of 98.68%, and an average total latency of 20.48 ms per query, outperforming traditional models such as XGBoost, LSTM, BERT, and SVM in both effectiveness and efficiency. The significance contribution of this work lies in providing a robust, scalable, and context-aware platform that ensures HIPAA-compliant, real-time medical guidance, enhancing patient care, reducing response delays, and enabling reliable AI-driven healthcare support.

Ähnliche Arbeiten