COVID-19 Analysis Hub

A web application providing interactive tools for exploring COVID-19 data and signals. The app offers two main features:

• Signal Correlation Analysis
• Forecasting

Live app: COVID-19 Analysis Hub

Overview

The COVID-19 Analysis Hub is designed to help public health professionals, epidemiologists, and researchers explore relationships between different COVID-19 signals and create forecasts. The application uses data from the Delphi Epidata API, maintained by Carnegie Mellon University.

Features

1. Signal Correlation Analysis

This tool allows users to:

• Compare any two COVID-19 signals (e.g., cases vs. deaths, hospitalizations vs. cases)
• Explore correlations at different geographic levels (nation, state, county, etc.)
• Calculate time-lagged correlations to identify leading/lagging relationships
• Choose between different correlation methods (Pearson, Kendall, Spearman)

The correlation analysis can help answer questions such as:

• How long after a rise in cases do we typically see a rise in hospitalizations?
• Which signals might serve as early warning indicators?
• How do relationships between signals vary across different regions?

2. Forecasting

The forecasting tool enables users to:

• Select multiple signals as predictors
• Choose the target signal to forecast
• Set the prediction date and forecast horizon
• Compare forecasts using different models:
  • ARX (AutoRegressive with eXogenous inputs)
  • Flatline
  • CDC Baseline

Key features include:

• Comparison between forecasts using real-time vs. revised data
• Confidence intervals for predictions
• Visual assessment of forecast accuracy
• Ability to experiment with different predictor combinations

Getting Started

1. Visit the live application
2. Choose either "Signal Correlation Analysis" or "Forecasting" from the home page
3. Follow the step-by-step instructions provided within each tool
4. Experiment with different signals, regions, and parameters

API Key (Optional)

The app works without an API key, but you may encounter rate limits. To avoid this:

1. Request a free API key from Delphi's Epidata here
2. Enter the key in the API Settings section on the home page

Feedback and Issues

Found a bug or have suggestions? Please let me know through our feedback form.

Technical Information

The application is built using:

• Streamlit for the web interface
• Python and R for data analysis
• Docker for containerization
• Google Cloud Run for hosting

The source code is available at GitHub.

Acknowledgments

All data is sourced from the Delphi Research Group at Carnegie Mellon University through their Epidata API. Signals integrated into the app include:

• Cases and deaths (from JHU CSSE)
• Hospitalizations (from HHS)
• Test positivity rates (from COVID Act Now)
• And others

Feel free to request more through the feedback form above. A full list of available signals can be found here.