Healthcare Chatbot for Slowing the Progression in Chronic Kidney Disease Stage 3 Patients

Abstract

Chronic kidney disease (CKD) is a significant global health problem characterized by the continuous deterioration of kidney function. It has been found that patients with stage 3 chronic kidney disease are at risk of progressing to end-stage kidney disease. This situation requires close medical attention to slow down kidney deterioration in these patients. To achieve this, a health chatbot system specifically designed for stage 3 CKD patients has been developed. The chatbot serves as a virtual representative of hospital staff, processing user messages based on medical knowledge stored in the system's database and responding in a chat format within the LINE application. Additionally, the chatbot's algorithm, which boasts very high accuracy, is vital for medical diagnosis. This research aims at the investigate of three techniques: Dialog flow, Gemini, and GPT3.5. These are powerful and widely used technique in natural language processing (NLP). We conducted a health chatbot in the outpatient department of internal medicine at Burapha University Hospital, Chon-buri Province, Thailand. The study sample consisted of 50 patients with stage 3 chronic kidney disease. These patients were divided into two groups: experimental group: 25 patients and control group: 25 patients. We classified the medical knowledge of chronic kidney based on insights from qualified medical experts which included 120 distinct sentences in the Thai language. Finally, the effectiveness of dialog flow, Gemini, and GPT3.5 in predicting CKD based on knowledge sentences was modeled. Our experiments compared the accuracies of Dialog flow, Gemini, and GPT3.5, which were found to be 88%, 75.61%, and 41.46%, respectively. As a result, Dialog flow outperformed the other methods.