The influence of meteorological factors on COVID-19 spread in Italy during the first and second wave

Abstract

The relation between meteorological factors and COVID-19 spread remains uncertain, particularly with regard to the role of temperature, relative humidity and solar ultraviolet (UV) radiation. To assess this relation, we investigated disease spread within Italy during 2020. The pandemic had a large and early impact in Italy, and during 2020 the effects of vaccination and viral variants had not yet complicated the dynamics. We used non-linear, spline-based Poisson regression of modeled temperature, UV and relative humidity, adjusting for mobility patterns and additional confounders, to estimate daily rates of COVID-19 new cases, hospital and intensive care unit admissions, and deaths during the two waves of the pandemic in Italy during 2020. We found little association between relative humidity and COVID-19 endpoints in both waves, whereas UV radiation above 40 kJ/m² showed a weak inverse association with hospital and ICU admissions in the first wave, and a stronger relation with all COVID-19 endpoints in the second wave. Temperature above 283 K (10 °C/50 °F) showed a strong non-linear negative relation with COVID-19 endpoints, with inconsistent relations below this cutpoint in the two waves. Given the biological plausibility of a relation between temperature and COVID-19, these data add support to the proposition that temperature above 283 K, and possibly high levels of solar UV radiation, reduced COVID-19 spread.

Keywords: COVID-19; Humidity; Meteorological factors; Temperature; Ultraviolet radiation.

Fig. 1

Daily COVID-19 endpoints and meteorological factors trends in Italy in 2020. The meteorological data were averaged over the provinces and modeled with restricted cubic splines with 7 knots.

Fig. 2

Average number of cases/hospitalizations/ICU/deaths per 100,000 inhabitants in the first wave, as a function of meteorological factors (temperature/UV radiation/relative humidity) estimated in random-effects restricted cubic spline Poisson regression models adjusting for population density, daily movements, old age index, number of single-family homes, PM_2.5 with a lag time of 1 day. The shaded area represents pointwise 95% confidence intervals.

Fig. 3

Average number of cases/hospitalizations/ICU/deaths per 100,000 inhabitants in the second wave, as a function of meteorological factors (temperature/UV radiation/relative humidity) estimated in random-effects restricted cubic spline Poisson regression models adjusting for population density, daily movements, number of single-family homes, old age index, PM_2.5 with a lag time of 1 day. The shaded area represents pointwise 95% confidence intervals.

Fig. 4

Akaike Information Criterion (AIC), when varying the lag time between health endpoints and meteorological variables, estimated in random-effects restricted cubic spline Poisson regression models. Lower values of AIC mean that the selected model is better. We adjusted for old age index, density, daily movements, old age index, number of single-family homes and PM_2.5, the latter with a lag time of 1 day. The dashed lines represent the lag times emerging from a priori interpretation based on nationwide epidemiologic data.