To represent the different regions of the U.S., we included four states—Connecticut, New York, California, and Texas—in the study. Generally, hotspots of LAN data were located in large cities or metro-centers for all four states tested, such as Hartford, CT, New York City, NY, Dallas, TX, and San Francisco, CA.
The maximum LAN intensity calculated by ArcMap tool was 254.68, 254.68, 235.14 and 190.30 in New York State, Connecticut, Texas, and California, respectively. The mean LAN intensity was 92.31 in Connecticut and around 40 in the other three states. The maximum COVID-19 case-rate was in Texas among all four states, which was around 67 cases/1k people, followed by California (around 47 cases/1k people), New York (around 43 cases/1k people), and lowest in Connecticut (around 27 cases/1k people) at the time of data collection.
For Connecticut and New York State, the hotspots of LAN intensity, COVID-19 case rates/1k during lockdown, and COVID-19 case rates/1k during the overall duration shared similar patterns that clustered around major cities in the state (Fig. 1), but were slightly different after reopening. Differently, in Texas and California, there were inconsistent patterns of hotspots for the four interested variables, but they share the trend that the hotspot-areas were similar for reopening and overall durations.
During the lockdown period, the hotspots of COVID-19 case rates shared similar patterns with LAN data in Connecticut and New York, but had different geo-patterns in California and Texas. During the reopening period, the locations of hotspots of COVID-19 case rates were very different from the LAN data map in New York, Texas, and California. The hotspot patterns of COVID-19 case rates during the overall period were similar to the LAN map in Connecticut and New York. On the contrary, the overall hotspots of COVID-19 case rates were similar to the reopening period in California and Texas.
An analysis of multivariable regression models also revealed similar patterns to the ones that geo-patterns showed. The cross comparison among four states showed that there were statistically significant correlations between LAN intensity and cases/1k for the overall and lockdown durations in New York and Connecticut (P< 0.001). There was no statistically significant association between LAN intensity and cases/1k, for the overall, lockdown, or reopening durations, in Texas and California (Table.1). The overall case rates were significantly associated with LAN in New York State (p<0.001) and Connecticut (p<0.001), in which every 1 unit increase of LAN had a 15.6% increase in the overall case rate in New York, and a 3.7% increase in Connecticut. The results of the lockdown period were similar to those for the overall period in New York(p<0.001) and Connecticut (p<0.001). During the reopening period, there was a significant small positive association between case rates and LAN data in Connecticut (p<0.001). Based on R-squared results, the state-specific regression models could explain more variations in New York (R2 = 0.80, 0.78 and 0.40) and Connecticut (R2=0.58, 0.57 and 0.21), compared to the data in California (R2 = 0.22, 0.37, 0.23) and Texas (R2 = 0.13, 0.02, 0.12), for the overall, lockdown and reopening durations.