Analysis code for: Phosphate-based refeeding syndrome definitions do not identify critically ill patients at higher mortality risk: a dual-cohort study of MIMIC-IV and the eICU-CRD
Year: 2026
Type: software
Abstract: SQL extraction queries and a Python analysis pipeline that reproduce the cohort construction and statistical results of a retrospective dual-cohort intensive-care study examining the post-nutrition serum phosphate trajectory and short-term mortality. The derivation cohort is drawn from MIMIC-IV (v3.1) and the external validation cohort from the eICU Collaborative Research Database (v2.0), both accessed through their Google BigQuery distributions on PhysioNet. The two SQL files build one analytic row per eligible ICU stay, applying the eligibility criteria (adults ≥18 years; first ICU admission; ICU stay ≥48 h; a baseline serum phosphate within 24 h before and a nadir within 72 h after nutrition initiation; baseline phosphate ≥2.0 mg/dL) and assembling the phosphate, magnesium and potassium trajectories, the caloric intake over the refeeding window, the covariate set, and the mortality outcomes. The Python script (Python 3.12.9; NumPy, SciPy, pandas) applies the phosphate plausibility filter, derives three phosphate-based refeeding-syndrome definitions (a traditional absolute-threshold definition, the ASPEN consensus definition, and the definition from the trial by Doig and colleagues), fits logistic-regression models by maximum likelihood with observed-information covariance, and produces the baseline table, the definition–mortality tables for both cohorts, the prespecified sensitivity and subgroup analyses, the quartile and electrolyte tables, the restricted-cubic-spline dose–response values, and the E-value for unmeasured confounding.
Authors: Shao, Lujun, Li, Yue, Tai, Huiyu, Wang, Li, Shi, Jinyu
Language: English
Subfield: Nutrition and Dietetics
Field: Nursing
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7166459209