Proposal of a Hybrid Clinical Decision Support System Based on Bayesian and Rule-Based Reasoning for Brazilian Primary Care

Clinical Decision Support Systems (CDSS) are pivotal for improving diagnostic accuracy and efficiency in Brazilian primary care. We propose a hybrid CDSS that integrates Bayesian reasoning—updating post-test probabilities via sequential likelihood‐ratio (LR) calculations—with rule-based inference for conditions governed by composite clinical criteria. Our database is implemented in SQLite and populated from systematic reviews and meta-analyses, ensuring each LR is grounded in high-quality evidence. An automated preprocessing pipeline normalizes free-text symptom inputs and maps them via a curated synonym table. To address the conditional independence assumption inherent in sequential LR application, we outline a roadmap toward Bayesian network extensions that explicitly model symptom interdependencies. Furthermore, we describe a two-stage validation plan: (1) retrospective evaluation on de-identified Brazilian patient records, and (2) prospective pilot testing with primary-care physicians in real-world consultations, assessing discrimination (AUC), calibration, usability, and time-to-diagnosis. This hybrid CDSS aims to deliver transparent, evidence-based differential diagnoses personalized to each patient’s profile, laying the groundwork for scalable, explainable decision support in Brazilian primary care.