Artificial intelligence models in primary care: performance, transparency, and safety in patient triage

Abstract

The incorporation of artificial intelligence (AI) into health systems has significantly progressed in recent years, expanding into non-hospital settings such as primary care and emergency services. This article critically analyzes international experiences (United States, Canada, United Kingdom, and Brazil) involving AI applications for automated triage, risk stratification, and clinical decision support, with a focus on low- and medium-complexity healthcare settings. It outlines key risks associated with these technologies—algorithmic bias, opacity, interoperability failures, data governance weaknesses, and privacy issues—in light of international regulatory and ethical frameworks proposed by institutions such as the World Health Organization (WHO), the Food and Drug Administration (FDA), and the National Institute for Health and Care Excellence (NICE). Based on this analysis, the article proposes a set of minimum criteria for the safe and ethical implementation of AI in primary care and emergency contexts, including local clinical validation, transparency, bias control, data governance, systemic integration, staff training, and post-deployment monitoring. It concludes that AI can strengthen primary care, improve patient flow management, and support complex clinical decisions, provided it is implemented under robust clinical governance, with continuous professional oversight and full respect for patients’ rights.