Effectiveness of Artificial Intelligence (AI) in the Prevention, Diagnosis, and Triage of Respiratory Diseases: A Multicenter, Randomized Controlled Study

Artificial intelligence (AI) technologies, particularly advanced large language models like GPT-4o developed by OpenAI, hold immense potential in enhancing public health education and preventive behaviors. Although GPT-4o was not specifically designed for respiratory disease prevention, it has shown promising prospects in numerous healthcare-related applications, such as providing health information, responding to public inquiries, and supporting health education efforts. However, its effectiveness in improving public awareness and management capabilities regarding respiratory diseases remains to be further explored.

Understanding and managing respiratory diseases involve complex processes, including symptom recognition, application of preventive knowledge, and informed decision-making. Integrating AI tools like GPT-4o into public health education could potentially enhance knowledge dissemination, reduce misinformation, and encourage appropriate preventive behaviors among the general population. Nevertheless, GPT-4o has not been specifically validated for respiratory disease prevention and carries the risk of generating misleading or inaccurate information, which could confuse users. Improper use of such tools may fail to raise awareness and could even lead to counterproductive behaviors. Therefore, studying how large language models like GPT-4o can effectively support public education and behavior change in this context is of critical importance.

In this study, participants will be randomly divided into two groups: one group will have access to Fine-turned GPT-4o, while the other will rely solely on traditional online tools. They will be presented with scenarios related to respiratory diseases and asked to explain their identification of high-risk factors, understanding of diagnoses, and proposed triage actions for each scenario. Each scenario was developed by a panel of three experts in respiratory health, who also established standardized answers. Responses will be evaluated by two independent groups of reviewers unaware of the participants' group assignments. These experts independently created initial scoring criteria and resolved discrepancies through multiple rounds of discussion to ensure consistency and accuracy.