ICU vital-sign record density: source-stream and site-level variation

Analysis code for a retrospective data-quality study of vital-sign record density across MIMIC-IV v3.1 and eICU-CRD v2.0. The study decomposes variance in documentation metrics by data stream, database, contributing site and care unit, and reports three consequences for the use of record density as a process measure: exposure composition includes shift-anchored alarm-limit records; an apparently architectural difference between databases is reproduced by the choice of source table; and a pooled low-density threshold designates hospitals rather than patients. No patient-level data are redistributed. Both source databases require credentialed access through PhysioNet.