Teaching Trajectory-Based Reasoning in Dengue: A Mentored LLM Workflow Anchored to WHO/CDC Guidance

Severe dengue can develop rapidly during a short critical phase, making early recognition of converging warning signs essential. This preprint describes a mentored, guideline-anchored educational workflow in which trainees use large language models (LLMs) under faculty supervision to practice trajectory-based reasoning from a de-identified dengue time-series (vitals, labs, and brief notes). Interpretation is anchored to WHO and CDC dengue guidance and reinforced through explicit verification steps and reproducibility artifacts (structured dataset, prompt scaffold, and an auditable workflow figure). The goal is training—standardizing how learners recognize and communicate evolving risk patterns—rather than automated diagnosis, with relevance to planetary health as climate change expands the burden of arboviral disease.