Discovery of a Latent Entropy-Based Physiologic State Variable Governing Multisystem Instability in Critical Illness

This dataset contains all code, intermediate tables, and outputs used to compute the Energy Index (EI), its time-derivative (EI_slope), and associated cross-correlation and regression analyses on patient vital signs. The files correspond exactly to Phases 03 and 04 of the computational workflow.

Contents:

notebook/ – A full Google Colab notebook exported as HTML and IPYNB, containing the exact code used for data cleaning, EI computation, EI_slope derivation, cross-correlation analysis, and multivariable regression.

tables/ – CSV files of the processed ei_full dataframe, the xcorr_df table (lag-based cross-correlation results), and the ss_df table (two-variable OLS models for EI_slope).

figures/ – Auto-generated plots of EI trajectories and regression diagnostics (if any were produced).

metadata/ – Variable dictionaries, version information for Python, NumPy, Pandas, Statsmodels, and notes on reproducibility.

Source data:
All raw clinical data originate from the MIMIC-IV database (PhysioNet) and are not redistributed here. Users must obtain their own access credentials to PhysioNet. This repository contains only derived variables and reproducible code.

File format:
ZIP archive. All files are UTF-8 encoded. Python 3.12 environment.

Purpose:
To allow full replication of the statistical pipeline for computing EI, EI_slope, and evaluating their associations with physiological variables in a real EHR dataset.