1/ Apparently @CDCDirector Dr. Robert Redfield is claiming that the spike in the #COVID-19 case growth rate in the South is _not_ due to premature relaxation of social distancing measures. This claim is directly contradicted by the evidence... https://twitter.com/ashishkjha/status/1283243422913040387

2/ In a study led by Mark Siedner ( @MGHMedicine @mgh_id @MonganInstitute) and Guy Harling ( @UCLGlobalHealth), we collected state-level data on social distancing measures. Original paper on @medrxiv, coming out in @plosmedicine soon: https://www.medrxiv.org/content/10.1101/2020.04.03.20052373v1?versioned=true

3/ We now supplement the data in our @PLOSMedicine paper with _new_ data on the _relaxation_ of social distancing measures. A new @medrxiv preprint will be posted here, and I will update this thread with a link when it is available.

4/ Our outcome is the state-specific effective reproduction number R(t), the expected number of secondary infections generated by each index case given date-specific population susceptibility and risk behavior. Helpfully provided by @cmmid_lshtm here: https://wellcomeopenresearch.org/articles/5-112 

5/ We fitted mixed-effects linear regression models with a random effect for state, specifying explanatory variables time in days, implementation period before vs. after relaxation, and a time-by-period interaction term

6/ Between April 20 and June 1, 2020, all 51 jurisdictions had relaxed at least one statewide social distancing measure. During the 3 wks prior to relaxation, R(t) was declining & at the time of initial relaxation, the median R(t) was 1. So far so good. (not really but whatever)

7/ After relaxation, R(t) starts going back up, by an average of 0.0047 units per day compared with the pre-relaxation period.

8/ Things look the same if you examine what happens after the relaxation of statewide restrictions on internal movement (ie, shelter-in-place orders, "lockdowns"), but their implementation was closely correlated with the initial relaxation of social distancing measures

9/ Lots of caveats, limitations on causal inference, voluntary restrictions on mobility, etc

10/10 These data are only current through May 28, but we are in the process of updating the data on social distancing measures and incorporating new estimates of R(t). Take a wild guess as to what things will look like when we extend the analysis to June...