Projections for first-wave COVID-19 deaths across the US using social-distancing measures derived from mobile phones

Spencer Woody; Mauricio Garcia Tec; Maytal Dahan; Kelly Gaither; Spencer Fox; Lauren Ancel Meyers; James G Scott

To construct predictors x for our regression model, we then take the first four principal components of the lagged social distancing metrics. These four components explain about 95% of the variation in the state-level data. Figure 2 : SafeGraph social-distancing data in New York versus Texas. The socialdistancing predictors we use in our model quantify visitation patterns to various points of interest, using GPS traces derived from mobile phones. The vertical axis is standardized so that 0 represents a pre-pandemic baseline, and -0.5 indicates a 50% decrease in visitations relative to that baseline. The vertical lines in each panel represent the day in which the state death rate reached 3 per 10 million residents. One can see that social distancing in Texas began substantially before this threshold day was reached, but much nearer this day in New York. 9 All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity.