See risk factors analyzed from one of the largest studies, OpenSafely, studying 17 million UK adults
Williamson, E.J., Walker, A.J., Bhaskaran, K. et al.
Factors associated with COVID-19-related death using OpenSAFELY.
Nature (2020). https://doi.org/10.1038/s41586-020-2521-4
Reported risk factors (age, male, chronic illness) agree with our initial analysis. Additional risk factors were obesity, being non-White and specific chronic conditions (e.g. immunosuppression, neurological, diabetes, prior transplant)
Many scientists have come together to share their data, code and ideas openly and timely to fight the COVID-19 epidemic (news link, free Alibaba cloud services) Here I've gathered and incorporated snippets from several publications and repos (see Acknowledgments) and also experimented with methods inspired by their work or other related fields (e.g. Social Network Analysis, gene expression networks, compartmental modeling).
We are based in SF Bay but collaborate remotely to solve this public health crisis. We come from diverse backgrounds ranging from Epidemiology, Public Health, Cheminformatics, Astrophysics, Mechanical Engineering.
- Github: https://github.com/yenlow/nCoV2019
- Slack channel: coronavirus2020.slack.com
- Google shared drive
Contact Yen Low to join or contribute code and ideas here in the Wiki or Issues!
- Downloads line lists from Google spreadsheet
- Sources R scripts to clean, preprocess and summarize data (from beoutbreakprepared/nCoV2019 and jameshay218/case_to_infection)
- Where possible, I tried to re-use their R code but also re-implemented them in Python to suit my needs and identify determinants of death cases.
As expected, risk factors are:
- chronic illness
- being in Wuhan
- increasing age
Epidemiology has always thrived on big data. Even back in the 1840s, John Snow, the father of Epidemiology, cleverly collected addresses from the local water utility provider and created the first outbreak dot map which located the Cholera epicenter to be the water pump on Broad St in London.
Today, we have even more data and methods available. This figure is a good way to see how they may all come together. With advances in deep sequential models, graph networks, stochastic agent-based modeling, etc, we can get really inventive!
