Optimal analysis methodology for tracking effect of BLM protests on subsequent covid rates I want to see if there is a correlation between BLM protests by time, place, and number of protesters with covid rates 1, 2, and 3 weeks from the date of the protest. The best methodology I could think of was essentially a case-control setup, where for each protest, I find a county of similar size, in the same state, with a similar case rate the time of protest, and compare fluctuations between those two counties.
Is there a more robust methodology to use? I know case-control are generally less favorable unless they're the only option.
Thanks for your help.
 A: A before-after comparison of positive viral tests per day, hospitalizations for COVID-19, and mortality due to COVID-19 within counties with BLM demonstrations is an alternative design.  It is not necessarily more robust but may be more practical.
An epidemiologist would not call the proposed study a case-control study.  In an epidemiologic case-control study, cases--people with disease--and controls--people without disease—are identified.  The cases and the controls are then categorized as exposed or unexposed to the factor of interest, looking backwards.
In the study described, there are counties where BLM demonstrations took place (exposed to BLM demonstrations) and counties where BLM demonstrations did not take place (unexposed to BLM demonstrations).  The proposed study would assess whether, going forward from the time of exposure or non-exposure to BLM demonstrations, infections with the SARS-CoV-2 virus and bad outcomes due spread of the virus (e.g., hospitalizations for COVID-19, death due to COVID-19) differ between the BLM exposed and BLM unexposed counties.  The study would be analogous to a prospective (cohort) study epidemiologically.
But a program evaluation paradigm fits better.  BLM demonstrations could be viewed as a “program” whose effect on viral spread is being evaluated by comparing counties that did and did not “participate” in the “program.”
The idea of a comparison group to assess the “effect” of exposure to BLM demonstrations with SARS-CoV-19 infections and outcomes at the county level is a good idea.  There are some practical problems.
Many of the places with BLM demonstrations were in very large cities—Seattle, LA, Denver, Phoenix, Miami, New York City, Detroit.  For these (and other) places, it seems unlikely that it will be possible to find unexposed counties of comparable size in the same state.  For example, there are no counties in Washington state that are close to King’s county (Seattle) in size, no counties in Arizona that are close to Maricopa county (Phoenix) in size, and no counties in California that are close to Los Angeles county  (LA) in size.  If the study idea is pursued, an approach to handling exposed counties with no comparable unexposed county should be specified in advance.
The study could potentially look at several different measures of viral spread—PCR-positive tests, hospitalizations for COVID-19, mortality due to COVID-19.  Care should be taken with the use of PCR-positive tests per day.  People attending BLM demonstrations were encouraged to “get tested.”  The number of positive tests per day might increase more in the counties exposed to BLM demonstrations because of greater use of testing, not because the virus was spread more quickly because of the BLM demonstrations.  That is, there is a potential bias due to greater test-seeking in counties exposed to BLM demonstrations because of encouragement to get tested.  Hospitalizations for COVID-19 and death due to COVID-19 are not susceptible to this bias.
The proposal to do a county-level comparison might be because data on positive tests for the virus, hospitalizations for COVID-19, and deaths due to COVID-19 are publicly available at the county level.  If city-level data are obtainable, a comparison of BLM exposed cities with BLM unexposed cities in the same state might make it possible to find comparators for a larger percentage of exposed places.
