GISSA GPT: An Agent-Oriented Architecture for Intelligent Governance in Digital Health

The GISSA system (Intelligent Governance in Health Systems) is a computing platform, implemented in 2020, that automatically collects data from Ministry of Health information systems in Brazil: e-SUS, CNES, SIM, SINASC, SI-PNI, and SINAN. It does so through data-extraction bots, analyzes these data using multiple intelligent technologies, and transforms them into integrated information that supports decision-making at different levels of health management. In 2021, the Networks and Systems Laboratory team at IFCE (LAR) implemented “Smart GISSA”, a health-governance system based on Machine Learning, defended as a doctoral dissertation at the Federal University of Ceará [9]. In 2024, the same LAR team developed “Giselle Saúde”, a sentiment-detection system using Generative AI in Digital Health [16]. This paper presents GISSA GPT, an evolution of Smart GISSA that incorporates Large Language Models (LLMs), leveraging the expertise acquired by the LAR team in building Giselle Saúde. It is an agent-oriented prototype for digital health governance with the following characteristics: (i) a generative-AI environment inspired by Smart GISSA; (ii) an event-driven design for integration among modules; and (iii) Retrieval-Augmented Generation (RAG) mechanisms to anchor responses in verifiable sources, implemented with modern technologies (n8n, LangChain, ChromaDB, etc.) and able to incorporate new protocols and guidelines selected by a multidisciplinary team [13, 18]. Considering that Giselle Saúde focuses on mental health, future work proposes integrating GISSA GPT with Giselle Saúde.