by Nicholas Archambault
The United States, a global leader in technology and innovation, has fared among the worst of all nations in mitigating the spread and impact of COVID-19. Cases are swelling toward a third peak as daily averages rise in four out of every five U.S. states, a resurgence driven by milder temperatures and a populace weary of the months of restrictions that social distancing and lockdown measures have imposed on their lives. Nine months into the pandemic, the country just recorded its worst week yet, with 528,927 new cases and a seven-day average of over 75,500, according to the New York Times. USA Today reports that a one-day record 88,521 cases were reported on October 29, equivalent to a new diagnosis every 0.976 seconds.
In addition to fighting a deadly pandemic, dealing with financial concerns and adhering to social distancing measures, the American public has been forced to confront a wave of disinformation regarding the virus and the proper steps our society must take to contain it. Science has become politicized to the extreme, in the moment we can least afford it to be. With an eye toward one of the most consequential and contentious elections in national memory, the president and other politicians have consistently downplayed the virus. Instead of formulating a sound response to the worst mass-casualty event in national history, they have prioritized undisguised political maneuvering to deepen the stark partisan divide among Americans over the costs and benefits of various public health measures.
Given the breadth of factors contributing to America's uniquely poor handling of the pandemic, any attempt to comprehensively analyze and track the spread of COVID-19 must consider the effects of influences beyond climbing case totals. This project integrates not just health data from the past six months, but also political preferences, social distancing and economic reopening procedures, and access to healthcare and resources in order to predict COVID-related deaths across the United States in the second week of November.
Data were collected from 11 diverse sources, substantially cleaned and manipulated, and combined into a dataframe that provides American states' individual totals not only for health-related statistics like positive cases and deaths, but also for engineered features such as population density, access to hospital beds per 100,000 civilians, and survey responses on mask-wearing practices.
Using this collection of traits and features germane to all the different angles of this pandemic, we attempt to predict state-by-state COVID-related deaths in the second week of November, 2020.
All documentation of data, description of processes, and explanatory plots can be found in the file, 'October Case Study'.