Hackathon: Public Health Informatics (Public Health Dynamics)

Submission Dates:
09/17/2021 to 10/03/2021
Citation Author(s):
Bobak
Mortazavi
Submitted by:
Sorush Omidvar
Last updated:
Sun, 10/03/2021 - 18:02
DOI:
10.21227/4vej-7w18
License:
Creative Commons Attribution
296 Views

Abstract 

Data Challenge Description:

COVID-19 has significantly impacted the entire world, changing the way of life for everyone. Knowing that the spread of the virus has been affecting almost all aspects of our day-to-day lives, it is critically important that we provide the relevant information, analyses, and predictions that can effectively inform the public so that everyone can make informed decisions during such epidemiological crisis.

Data Challenge Goal:

The goal of this data challenge is to predict the 7-day average of new COVID-19 cases and that of the positivity rate based on the historical public health data. Accurate prediction of such epidemiological trends can provide useful insights for the public, helping them make informed decisions regarding protection/mitigation measures, travel planning, and others.

Instructions: 

Datasets:

Covid Act Now (CAN) (https://covidactnow.org/about) is the COVID-focused initiative of the Act Now Coalition, whose COVID data and risk assessment include every U.S. State, 380+ metros, 3,100+ counties, and two U.S. territories. So far, CAN have served over 15m website users since the beginning of the pandemic, and kept more than 200k subscribers up to date on COVID news and alerts enabling informed decisions for individuals to cope with the pandemic. CAN has worked with more than 100 federal, state, and county officials as well as numerous multinational corporations and NGOs to develop data-driven COVID responses. CAN provides public access to the latest epidemiological data related to COVID-19, where the data comes from official sources including the U.S. Department of Health and Human Services, the Centers for Disease Control and Prevention, The New York Times, and official state and county dashboards. They provide an API for accessing the data, along with data definitions for all fields and an overview of how to query data for states, counties, or metros. CAN’s data areupdated daily around noon Eastern Time, where updates reflect the latest available data from the official sources stated above.

In addition to the epidemiological data provided by CAN API, participants are welcome to consider additional data (such as Google mobility data) that may be helpful in more accurate and reliable prediction of future trends.

Download link:

  • The COVID-19 epidemiological data can be downloaded from the Covid Act Now (CAN) official website through the CAN Data API: https://covidactnow.org/data-api
  • Registration is need to get access to the dataset, after which the registered user will have immediate access to the historical and the latest COVID-19 epidemiological data.
  • Google mobility data can be obtained from: https://www.google.com/covid19/mobility/

Evaluation criteria:

Participants should use the historical epidemiological data until the report submission deadline (Sep. 26, 2021). In their report, they should include predictions for the 1-week prediction period (Sep. 27, 2021 – Oct. 4, 2021) following the submission deadline.

All participants will need to submit their predictions for the following values:

  1. Daily new cases for the 1-week prediction period
  2. 7-day average of daily new cases for every 7-day window ending on each day of the1-week prediction period
  3. Daily positivity rate for the 1-week prediction period
  4. 7-day average of the positivity rate for every 7-day window ending on each day of the1-week prediction period

Along with these future predictions, all participants should report how accurately their trained models were able to fit the historical data for the 4-week period ending on the submission date (Aug. 30, 2021 – Sep. 26, 2021).

Submitted predictions will be evaluated based on a combination of the following metrics:

  1. How accurately is the predictive model able to model the trends in the past?
  2. How accurately can the model predict future trends for the next 1-week period?

Submission:

    1. Technical description of your method
    2. Predictions for the period Sep.27,2021 – Oct. 4., 2021 (as described in the above evaluation criteria)
    3. Evaluation results of how accurately your trained model was able to fit the historical data for the 4-week period (Aug. 30, 2021 – Sep. 26, 2021), in terms of RMSE (root mean square error) and MAE (mean absolute error).
    4. URL to your GitHub repo, where your code can be accessed and the prediction results can be reproduced.

References:

Shah, Nirav R., Debbie Lai, and C. Jason Wang. “An impact-oriented approach to epidemiologicalmodeling.” Journal of General Internal Medicine 36.6 (2021): 1765-1767.