Project Author: Bruno Ferreira
Data Source: MIMIC-IV Demo Dataset
In this project, I explore the Medical Information Mart for Intensive Care (MIMIC)-IV demo dataset, a valuable resource comprising de-identified electronic health records (EHR) from 100 patients admitted to the Beth Israel Deaconess Medical Center between 2011-2016, all from emergency departments or intensive care units. MIMIC-IV can serve as a foundational repository for healthcare research, enabling exploration of diverse clinical scenarios, predictive modeling, and risk stratification.
This project presents a comprehensive analysis designed to extract meaningful insights from the MIMIC-IV demo dataset. Leveraging Python libraries and industry-standard methodologies, we'll aim to uncover trends, patterns, and create predictive models for patient outcomes and risk assessment.
MIMIC-IV is grouped into two primary modules: 'hosp' and 'icu'. The 'hosp' module encompasses data derived from the hospital-wide electronic health record (EHR), while the 'icu' module provides data from the ICU specific clinical information system at the Beth Israel Deaconess Medical Center. Key components of the dataset include patient demographics, miscellaneous health measurements, hospitalizations, diagnoses, laboratory measurements, medication administration, clinical procedures, and more.
The project will be divided into 3 subprojects, with the following goals:
- Cleaning and preprocessing the raw data into a more useful format (Data Cleaning & Preparation Notebook).
- Explore patterns and trends in critical care patient data (Exploratory Data Analysis Notebook).
- Create predictive model for in-hospital mortality (Predicting Hospital Mortality Notebook).