[ScreenGPT - The opportunities and limitations of artificial intelligence in primary, secondary and tertiary...

Introduction: Prevention and healthcare screenings are gaining popularity. Empowered patients, driven by curiosity and growing awareness, seek online healthcare information, regardless of its reliability. ChatGPT has simplified this process, motivating people to use it to ask healthcare-related questions, manage their healthy lifestyle, and even for self-diagnosis. Despite the fact that this tool cannot replace consultations with healthcare professionals, it could serve as a complement to traditional prevention processes. Objective: The objective of our research was to identify the fields where ChatGPT can effectively be used for primary, secondary and tertiary prevention. Additionally, we aimed to create a concept for an artificial intelligence-based service that could assist patients at various levels of prevention. Method: ChatGPT was analyzed and tested to determine its applicability at the three levels of prevention. Based on these capabilities, we used Python programming language to create the concept of new services, relying on the GPT-4 model. To increase the accuracy of responses, we used structured prompts. The application was made available and testable through the cloud service of Streamlit framework. Results: The tests identified several areas where the capabilities of ChatGPT could be utilized. Based on the results, we successfully established the foundations of a new service called ScreenGPT. Conclusion: We have ascertained that ChatGPT can provide useful answers to precise questions at all three levels of prevention. Although its responses accurately reflect human conversation, it relies on statistical methods for answer generation, so it is important for users to critically evaluate its answers. Based on these experiences, we have been able to make the ScreenGPT service available, however, numerous further investigations and work are needed to increase its reliability. Orv Hetil. 2024; 165(16): 629–635.