GC028-0004
The Dynamics of Covid-19: Weather, Demographics and Infection Timeline

Tuesday, 8 December 2020
Poster
Renato Pedrosa, UNICAMP State University of Campinas, Department of Science and Technology Policy, Campinas, Brazil
Abstract:
Background

Associations between daily growth rate of cases or deaths of COVID-19 and weather variables have been expected and reported by various studies. However, the effects of infection timeline or population density have not been accounted for so far, which we do in this paper.

Methods and results

Employing detailed data about weather (temperature and absolute humidity) for US states, the onset date of Covid-19 infection (date of 25th death) and population density, we did not find statistically significant associations between weather variables and the daily growth rate of deaths for U.S. states, but strong negative association with disease onset date (the later the disease started, the lower the growth rate) and positive one with higher density of counties where the disease first hit. In the case of countries, they both showed negative association, but once date of 25th death, the infection timeline variable considered, is introduced as regressor, weather variables become non-significant. The model employing the timeline variable and population density (R2=0.84) indicated that, for each 10 days added to the date of 25th death, there is a reduction of 0.054 point (0.041 to 0.067, 95% CI), while doubling the population density would add 0.015 point (0.0082 to 0.0214, 95% CI), to the daily growth rate of deaths by COVID-19. For countries, adding 10 days to the timeline variable would have an associated increase of 0.049 point (0.039 to 0.060, 95%CI) to the early daily growth rate of deaths.

Conclusions

Our results suggest that weather is not a determinant of the early growth rate of COVID-19, if one takes into account the timeline of the disease’s outbreak, for countries or U.S. states, likely due to early containment measures. Population density is positively associated to early pace of infection and its effects are better detected when employed in sub regional contexts. Our results, developed from data from March and April 2020, have been confirmed by continuing Covid-19 infections in the US during the Summer months, especially in warm and humid weather states in the Southeast and Southwes, like Florida and Texas. The same applies to Brazil and other South American countries, from the beginning of the pandemic.