suspicion-of-infection

Identify suspected infection events in MIMIC-IV by linking antibiotic administration records to culture collection timings. This Skill matches antibiotics to cultures within a 72-hour pre-antibiotic and 24-hour post-antibiotic window to support Sepsis-3 cohort construction.

The time window for detecting suspected infections links antibiotic administrations to cultures collected up to 72 hours before and 24 hours after the antibiotic was given. This defined window ensures accurate temporal matching for infection onset research.

You can use this suspicion of infection logic with eICU datasets as well as MIMIC-IV. It provides dataset-agnostic guidance for implementing the antibiotic and culture matching rules across similar ICU datasets.

Handle non-positive cultures and one-row-per-antibiotic matching during clinical data derivation by enforcing specific culture-before and culture-after rules. The process ensures each antibiotic entry is matched correctly without duplicating records for non-positive culture results.

This approach supports antibiotic stewardship studies and time-to-treatment analyses in ICU datasets by accurately timestamping suspected infection events. Linking antibiotic administrations to culture collections provides the temporal data needed for these clinical research workflows.

Use antibiotic administration timing instead of diagnosis codes to identify suspected infections because medication and culture data reflect real-time clinical decision-making. This Sepsis-3 aligned method captures actual suspected infection events rather than relying on retrospective billing codes.