Abstract
Progress notes are among the most clinically meaningful artifacts in anElectronic Health Record (EHR), offering temporally grounded insights into apatient's evolving condition, treatments, and care decisions. Despite theirimportance, they are severely underrepresented in large-scale EHR datasets. Forinstance, in the widely used Medical Information Mart for Intensive Care III(MIMIC-III) dataset, only about $8.56\%$ of hospital visits include progressnotes, leaving gaps in longitudinal patient narratives. In contrast, thedataset contains a diverse array of other note types, each capturing differentaspects of care. We present DENSE (Documenting Evolving Progress Notes from ScatteredEvidence), a system designed to align with clinical documentation workflows bysimulating how physicians reference past encounters while drafting progressnotes. The system introduces a fine-grained note categorization and a temporalalignment mechanism that organizes heterogeneous notes across visits intostructured, chronological inputs. At its core, DENSE leverages a clinicallyinformed retrieval strategy to identify temporally and semantically relevantcontent from both current and prior visits. This retrieved evidence is used toprompt a large language model (LLM) to generate clinically coherent andtemporally aware progress notes. We evaluate DENSE on a curated cohort of patients with multiple visits andcomplete progress note documentation. The generated notes demonstrate stronglongitudinal fidelity, achieving a temporal alignment ratio of $1.089$,surpassing the continuity observed in original notes. By restoring narrativecoherence across fragmented documentation, our system supports improveddownstream tasks such as summarization, predictive modeling, and clinicaldecision support, offering a scalable solution for LLM-driven note synthesis inreal-world healthcare settings.