MIMIC-IV hourly-adjusted urine output and AKI analysis

SQL queries and R workflow to create datasets for hourly-adjusted urine output (UO) and KDIGO staging as well as the original research results from the MIMIC critical care database. This repository accompanies the article: "Toward the standardization of big datasets of urine output for AKI analysis: a multicenter validation study".

Table of contents

• Introduction
• Main Objectives and Structure
• Requirements and Compatability
• Example
• Citation

Introduction

Accurate diagnosis and analysis of oliguric-AKI relies on timely UO charting. The lack of standardization in handling UO data and the various interpretations of KDIGO-UO guidelines limit the ability to make consistent comparisons and draw general conclusions. We aimed to establish a method for standardizing hourly UO using real-life charting data and to examine whether this method can identify oliguric-AKI. We also aimed to validate the method externally.

The model described, based on simple charting, can be used across the board for oliguric-AKI research. It may serve to analyze publicly available DBs and data sourced from standard EHRs as well as custom-made data in Excel tables.

This repository addresses the derivation cohort.

For the validation cohort see: https://github.com/arielhasidim/aumc-uo-and-aki

Main Objectives and Structure

This repository has two main objectives:

1. Enable the creation of hourly-adjusted UO and AKI events tables - For further instructions visit: create_data/....

2. Enable the reproduction of the associtated article - For further instructions visit: reproduce_article/....

Requirements and Compatability

• The MIMIC-IV database is required in order to run this code and is not provided with this repository.
• To access the MIMIC database, you will need to:
  • Become a credentialed user on PhysioNet and sign the use agreement (see 'Getting Started' tutorial).
  • Have MIMIC BigQuery (cloud) access (see 'Getting Started/Cloud' tutorial )
• You will need a Google Cloud Platform (GCS) billing account to run the queries.
• The SQL queries are written in GoogleSQL dialect (formally known as "Standard-SQL" dialect) and are probably compatible with other common dialects.
• The code was tested on MIMIC-IV v2.2 and MIMIC Code v2.4.0.

Example

After creating all the tables and reproducing the associated study, you should end up with a result page in HTML format: https://arielhasidim.github.io/mimic-uo-and-aki.

Citation

If you use this repository, please cite:

Hasidim, A.A., Klein, M.A., Ben Shitrit, I. et al. Toward the standardization of big datasets of urine output for AKI analysis: a multicenter validation study. Sci Rep 15, 20009 (2025). https://doi.org/10.1038/s41598-025-95535-4

@article{Hasidim2025,
  author  = {Hasidim, Ariel Avraham and Klein, Matthew Adam and Ben Shitrit, Itamar and Einav, Sharon and Ilan, Karny and Fuchs, Lior},
  title   = {Toward the standardization of big datasets of urine output for AKI analysis: a multicenter validation study},
  journal = {Scientific Reports},
  year    = {2025},
  volume  = {15},
  number  = {1},
  pages   = {20009},
  doi     = {10.1038/s41598-025-95535-4},
  url     = {https://doi.org/10.1038/s41598-025-95535-4}
}