Classifying Prolonged Invasive Ventilation in Elderly ICU Patients

Author: Max Vo
Machine Learning Analysis of MIMIC-IV (2026)
Repository: https://github.com/maxxhvo/Invasive_Ventilation_ML

🔐 BigQuery Authentication

A .json service account token is required to access the BigQuery instance containing the MIMIC-IV data.

• Place the .json token in the project root directory
• This token enables authentication and ensures reproducible access to the underlying data

Note: Do not commit this file to GitHub. Add it to your .gitignore.

---

🔁 Reproducibility Workflow

To reproduce the full project pipeline, run the Quarto (.qmd) files in the following order:

1. ~/02_Data_Wrangling/DataWrangling.qmd
2. ~/03_Pre-Processing+Feature_Engineering/Pre-Processing+FeatureEngineering.qmd
3. ~/04_Models+ML/Modeling.qmd

Each stage builds on the previous one and generates the datasets and models used in the final application.

---

Running the Shiny App

To launch the application locally:

# Replace <path-to-repo> with the location where you cloned this repository
setwd("<path-to-repo>/Shiny_Invasive_Ventilation")
shiny::runApp()

Or Run on a Posit IDE

---

📊 About the Application

This Shiny app provides an interactive interface for exploring and modeling prolonged invasive ventilation using MIMIC-IV data.

Key features include:

• Cohort exploration and visualization
• Patient-level ventilation timelines
• Raw data browsing
• Machine learning–based probability prediction for prolonged ventilation (> 2 days)

Data Source: This project uses the MIMIC-IV clinical database, a large, publicly available dataset of de-identified ICU patient records.