Statistical analysis

For statistical analysis, the study focused on four 14-day periods in 2020, which were compared to the corresponding periods in 2019: 31st March–13th April (climax of the first wave), 16th–29th June (early post-wave), 14th–27th July (mid post-wave) and 18th–31st August (late post-wave period). The last three study periods were chosen a priori to be evenly distributed, allowing two weeks of adaptation after withdrawal of lockdown measures.

Count data were expressed with absolute number and proportion. Using the Poisson regression, we estimated the percent change and its 95% confidence interval (CI) from the exponentiated Poisson regression coefficient. Type-III P values were used to assess whether the Poisson regression model with a specific variable was statistically significant (P value < 0.05). Data were displayed using locally estimated scatterplot smoothing (LOESS), in order better show data trend (smoothing span conservatively set at 14%). Extraction of count data and graphs were done with Microsoft Excel (Microsoft Corp., ver. 16.0), MedCalc (MedCalc Software Ltd, ver. 19.5.2), and all statistical analyses were performed using SPSS (IBM Corp., ver. 25.0).