Hospital Readmission Analysis

Project Overview

Hospital readmissions within 30 days after discharge are a major challenge for modern healthcare. They increase hospital costs and can lead to penalties under programs like the Hospital Readmission Reduction Program (HRRP).

This project uses a data-driven approach to identify patients at higher risk of readmission and supports proactive decision-making. Using patient history data, we analyze key factors that contribute to readmission within 30 days and provide actionable insights for hospital care.

Live Demo

You can view the live dashboard here:
🔗 Streamlit App

Project Documentation

Full project report can be found here:
📄 Project PDF

Project Screenshots

1. Dashboard Home

2. Age Distribution of Patients

3. Readmission Rate by Medication Type

4. 30-Day Readmission Rate by VCI Risk Group

Objective

• Analyze hospital patient data to understand patterns behind 30-day readmissions.
• Identify high-risk patients using clinical and operational features.
• Use the Vitality Complexity Index (VCI) and other patient metrics to support data-driven decisions.

Tools & Technologies

• Python 3
• Pandas & NumPy for data cleaning and processing
• Plotly Express for interactive visualizations
• Streamlit for building the dashboard

Data Description

The dataset contains patient-level data including:

• Demographics (age, gender, race)
• Admission details (emergency or planned, length of stay)
• Clinical and complexity data (number of diagnoses, lab procedures, VCI score)
• Medication and treatment data (medication type, changes, insulin/oral)
• Discharge disposition

Project Workflow

1. Data Collection and Cleaning

   • Combined multiple CSVs into one main dataframe
   • Standardized nulls, updated column types, created meaningful columns

2. Exploratory Data Analysis

   • Examined demographic, clinical, operational, and medication features
   • Created charts and visualizations for better insights

3. Feature Engineering

   • Created readmit_30 for 30-day readmission
   • Calculated scores like L_score, A_score, C_score, E_score
   • Combined scores to create the VCI Score and risk categories

4. Insights & Key Findings

   • Readmissions are low but critical
   • Gender and race have minimal effect on readmission (<3%)
   • Medication changes during hospitalization increase readmission risk (~12%)
   • Emergency admissions are linked to higher early readmission
   • High VCI patients are the strongest predictors of readmission

5. Recommendations

   • Focus attention on high-risk patients (high VCI scores)
   • Adjust nurse staffing based on patient complexity
   • Provide extra care for patients discharged to nursing homes
   • Use data-driven decisions to reduce readmissions and hospital costs

Dashboard Visualizations

The Streamlit dashboard includes:

• Age distribution of patients
• Readmission status distribution
• Readmission rates by race, gender, medication type, and medication changes
• Scatter plot of hospital stay vs lab procedures
• Correlation heatmap of numerical features
• Box plots of stay time by readmission status
• 30-day readmission rate by discharge disposition
• Readmission rates by VCI Risk Group