Fig. 1: Flow diagram of the cohort processing steps.
From: DRG-LLaMA : tuning LLaMA model to predict diagnosis-related group for hospitalized patients
We used regular expressions to extract the “brief hospital course” section from discharge summaries in MIMIC-IV dataset as input text. We filtered the discharge summaries that were of low quality, identified by either duplicated content or containing less than 40 words. We focused on MS-DRGs and consolidated all MS-DRG codes to version 34.0. Additionally, we filtered out rare DRGs with less than 2 occurrences in the cohort.
