Environmental factors and mobility predict COVID-19 seasonality

Background

We recently showed that seasonal patterns of COVID-19 incidence and Influenza-Like Illnesses incidence are highly similar, in a country in the temperate climate zone, such as the Netherlands (latitude: 52 ^° N). We hypothesize that in The Netherlands the same environmental factors and mobility trends that are associated with the seasonality of flu-like illnesses are predictors of COVID-19 seasonality as well.

Methods

We used meteorological, pollen/hay fever and mobility data from the Netherlands with its 17.4 million inhabitants. For the reproduction number of COVID-19 (R _t ), we used data from the Dutch State Institute for Public Health. This R _t metric is a daily estimate that is based on positive COVID-19 tests in the Netherlands in hospitals and municipalities. For all datasets we selected the overlapping period of COVID-19 and the first allergy season: from February 17, 2020 till September 21, 2020 (total number of measurements: n = 218), the end of pollen season. Backward stepwise multiple linear regression was used to develop an environmental prediction model of the R _t of COVID-19. Next, we studied whether adding mobility trends to an environmental model improved the predictive power.

Results

By means of stepwise backward multiple linear regression four highly significant (p value < 0.01) predictive factors are selected in our combined model: temperature, solar radiation, hay fever incidence, and mobility to indoor recreation locations. Our combined model explains 87.5% of the variance of R _t of COVID-19 and has a good and highly significant fit: F(4, 213) = 374.2, p-value < 0.00001. The combined model had a better overall predictive performance compared to a solely environmental model, which still explains 77.3% of the variance of R _t , and a good and highly significant fit: F _{(4, 213)} = 181.3, p < 0.00001.

Conclusions

We conclude that the combined mobility and environmental model can adequately predict the seasonality of COVID-19 in a country with a temperate climate like the Netherlands. In this model higher solar radiation, higher temperature and hay fever are related to lower COVID-19 reproduction, and mobility to indoor recreation locations with increased COVID-19 spread.

Highlights

• The seasonality of COVID-19 can be well-explained by environmental factors and mobility.

• A combined model explains 87.5% of the variance of the reproduction number of COVID-19

• Inhibitors of the reproduction number of COVID-19 are higher solar radiation, and seasonal allergens/allergies.

• Mobility, especially to indoor recreation locations, increases the reproduction number of COVID-19.

• Temperature has no direct effect on the reproduction number of COVID-19, but affects mobility and seasonal allergens.

• Adding mobility trends to an environmental model improves the predictive value regarding the reproduction number of COVID-19.